A New Technique for High-Throughput Solubility Assay

A New Technique for High-Throughput Solubility Assay P.E. Nielsen, C. Du, K. Tsinman, D. Voloboy, A. Avdeef pION INC., 5 Constitution Way, Woburn, MA 01801-1024, USA; [email protected] OBJECTIVES

METHODS

RESULTS

Compare two new, high speed, UV spectroscopy methods for drug solubility determination to two current methods: Saturation ShakeFlask and the ‘gold standard’ pSOL-3 potentiometric titration.

The robotic sample preparation system of the PSR4s Solubility Analyzer is similar to that of the PSR4p Permeability Analyzer (described elsewhere). It is based on the TECAN Genesis 100/4 worktable. The UV scanning spectrophotometer used was the Molecular Devices SPECTRAmax 190 for 96 well microtitre plates. The scan range was 190 nm to 500 nm. Plastic UV plates were used and as a consequence, only data above 240 nm were included in the analysis.

The solubility of a number of compounds that had been measured in different ways were compared to the high-throughput, microtitre plate based results from the PSR4s. Data from the aqueous dilution method had been reported before, but the data from the cosolventmethod are new. Table 1 shows the comparative results.

Present an improved method for determining concentration by UV spectrophotometry. Suggest a QC-scheme for the high speed methods centered around a set of standard drugs. Show how pH-mapping can be used to create solubility-pH profiles.

INTRODUCTION In order for any drug to be transported through the membranes of the human body, the drug molecule must be dissolvable in the aqueous phase of the intestinal fluid. Without dissolution, the drug would pass through the GI-tract as would brick-dust. For ionizable drug molecules, aqueous solubility is dependent on pH. When a solute molecule, weak acid (HA) or weak base (B), is in equilibrium with its precipitated form (HA(s) or B(s)) the equilibrium constant is defined as: (1)

S0 = [HA]/[HA(s)] = [HA]

or

S0 = [B]/[B(s)] = [B]

since by convention, [HA(s)] = 1 and [B(s)] = 1. The zero subscript denotes the zero charge of the precipitating species.

To determine the solubility of an unknown compound requires a reference of known compound concentration. A precipitation-free reference sample must be produced based on the known stock compound concentration. Two different methods may be used to accomplish this for determination of solubility of compounds, the properties of which are unknown a priori.

The aqueous dilution method adds a known, large amount of buffer to the sample to avoid precipitation in the reference sample. The UV spectrum of the reference is read in small increments from 190 nm to 500 nm. Mathematical treatment of the spectral data produces the areaunder-the-curve, AUCR, of the reference solution. The sample is prepared by adding a known quantity of stock sample to a known volume of a universal buffer solution of known pH. The quantity of sample must be X times higher than for the reference solution above, and it must cause precipitation to occur in the formed saturated solution. After a waiting period to let the saturated solution reach steady state, it is filtered to remove the precipitate. The UV spectrum of the sample is read in small increments from 190 nm to 500 nm. Mathematical treatment of the spectral data produces the areaunder-the-curve, AUCS, of the filtered solution. The ratio of the two numbers above, R = AUCR/AUCS, is used to recognize that the right conditions for solubility determination are present: reference has no precipitate, and the sample solution is saturated. Under these conditions, solubility can be determined as

For a monoprotic acid, the dissociation equilibrium must be considered as well: (8) (2)

H+ + A− = HA and

Where CR is the calculated concentration of the reference solution

S = CR/R

K1 = [HA]/[H+][A−] The results of a pH 4.0 − 9.5 assay are shown in Fig. 2.

In a saturated solution, solubility, S, at a particular pH is defined as the sum of the concentrations of all the species dissolved in the aqueous solution: (3)

S = [A−] + [HA]

Substitution of (1) and (2) into (3) produces the equation: (4)

S = S0(1+10±(pKa−pH))

Where ‘–’ of the ‘±’ sign applies to acids and ‘+’ to bases.

For an ampholyte, both signs must be observed: (5)

S = S0(1+10−(pKa1−pH) +10+(pKa2−pH))

Where pKa1< pKa2.

For multiprotic acids or bases, this equation applies (triprotic case): (6)

S = S0(1+(1+(1+10±∆pH1)10±∆pH2 )10±∆pH3 )

Where ‘–’ of the ‘±’ sign applies to acids, ‘+’ to bases and ∆pHn = pKa – n pH

The cosolvent method uses a suitable cosolvent to prevent precipitation in the high concentration reference solution. A known quantity of sample is added to a known volume of a universal buffer of known pH. A volume Y of cosolvent is added to volume Z of the reference solution. In this new solution the compound has been diluted by Z/(Y+Z). The UV spectrum of the reference is read in small increments from 190 nm to 500 nm. Mathematical treatment of the spectral data produces the area-under-the-curve, AUCRCOS, of the cosolvent reference solution. The sample is prepared as for the aqueous dilution method using a quantity of sample comparable to the amount used for the reference solution. After the waiting period and sample filtration, just before the UV spectra are read, a volume Y of water-miscible cosolvent is added to volume Z of the sample solution to produce a new solution in which the sample is now diluted by Z/(Y+Z). The cosolvent must have high solubilizing power, low vapor pressure, and low UV absorption. The UV spectrum of the sample is read in small increments from 190 nm to 500 nm. Mathematical treatment of the spectral data produces the areaunder-the-curve, AUCSCOS, of the filtered cosolvent sample solution.

A complete discussion of these equations is given in [1].

The solubility of the sample compound is then computed as:

While drug solubility is often measured at one pH, the pH-gradient of the GI-tract has a range of 1.5 to about 8. The concentration of the drug varies considerably along the way, if the compound is ionizable. The availability of a solubility-pH profile may therefore be helpful in evaluating the properties of the compound. Such profiles have been measured and reported before [2] but have been time consuming to produce. The high-throughput methods described here are able to automatically produce 8 or 16 solubility-pH profiles in one experiment.

(9)

Most often compounds are presented as a stock solution, 10 mM or more in DMSO. Even when this stock solution is diluted with buffer to a high degree, the remaining DMSO can significantly increase the measured solubility. If the pKa of the compound is known, the true value of S0 can be determined as shown in the log-log graph, Fig. 1. The weak base chlorpromazine has one pKa at 9.24, but when the profile is measured, it appears to be at 8.50. A shift, ∆, of 0.74 to the left. An acid under the same conditions would show a shift to the right. The log S0- value would be shifted higher by the same amount: (7)

S0apparent = S0 + ∆

-1

S = (1+Y/Z) CRCOS/RCOS

Measured profile

log S0app

When unknown samples are processed on the system, occasionally the whole set of data may appear out of line with expectations. To ensure that equipment problems are not to blame, it is advantageous to include one or two ‘standard’ compounds in the set of unknowns. The green-marked compounds in Table 1 represent good choices for such standards.

∆

-5 ∆ pK aapp pKa

8

9

REFERENCES

logS0

10

11

pH Fig. 1 The measured solubility-pH profile is shifted left as indicated by the difference between the true pKa and the apparent pKaapp. If known, this shift can be used to correct the determined S0app to obtain S0.

6/10/01 PN

Corr. S0 (µg/mL)

amitriptyline

B

9.45 a

56.9

3

chlorpromazine

B

9.24 a

19.4

3.4

diclofenac

HA

3.00 b

22.6

3.8

flurbiprofen

HA

4.03

furosemide

H2A

10.63, 3.52 b

29.8

griseofulvin

cos-S0app (µg/mL)

k

m

2.0 a

2.0 a

3.5 a

0.1 a

5.0

0.8 b

0.6 b

5.9 b

12.0 b, (2.9 c)

40.1 2.9

n

n/a

37.6

4.42 a

7.2

4.1

B

6.07 f

11.1

1.6

2-naphthoic acid

HA

4.16 f

33.3

20.2

15.0

phenazopyridine

B

5.15 f

12.2

12.2

24.5

piroxicam

XH

5.07, 2.33 h

10.5

1.1

22.1

probenecid

HA

3.01 f

4.6

0.7

7.4

terfenadine

B

9.53 f

4.4

0.1

miconazole

ShakeFlask S0 (µg/mL)

1.1

HA

indomethacin

pSOL-3 S0 (µg/mL)

9d

27.6 5.3

2.0 a

2.0 a, 1e

0.7 f 22.4 g 14.3 f 9.1i (3.3 c), 8-16 j (2.2-4.4 c) 0.6 f 0.1 f

verapamil B 9.07 4.4 0.1 53.1 0.1 f a M.A. Strafford A. Avdeef, P. Arthursson, C.A.S. Johansson, K. Luthman, C.R. Brownell, R. Lyon. Am. Assoc. Pharm. Sci. Ann. Mtng 2000, Poster presentation. b J.H.G. Jonkman, C.A. Hunt. Pharm Weekblad Sci. Ed. 1983, 5, 41-47. c Corrected for aggregate formation. Unpublished data. d J. Huskonen, M. Salo, J. Taskinen. J. Chem. Int. Comp. Soc.. 1998, 38, 450-456. e Y. Gur, I. Ravina, A.J. Babchin. On the electrical double layer theory. II. The PoissonBoltzman equation including hydration forces. J. Colloid Inter. Sci. 1978, 64, 333-341. f pION, unpublished data. g K.G. Money, M.A. Mintun, K.J. Himmelstein, V.J. Stella. J. Pharm Sci. 1981, 70, 13-22. h A. Avdeef, K.J. Box. Sirius Technical Application Notes (STAN). Vol 2, 1995. i C.R. Brownell, FDA, private correspondence, 2000. j J. Jinno, D.M. Oh, J.R. Crison, G.L. Amidon. Dissolution of ionizable water-insoluble drugs: the combined effect of pH and surfactant. J. Pharm. Sci. 2000, 89, 268-274. k A. Avdeef. Physicochemical Profiling. Current Topics in Medicinal Chemistry. (Submitted). m pION, unpublished data, DMSO concentration in sample 0.26% v/v. 2001.

The output created by the system takes many different forms from data tables to 3-D graphs that can be rotated under mouse control. If solubility-pH profiles are requested, graphs like the simulated one in Fig. 3 will be shown. Fig. 3 The solubility-pH profile for chlorpromazine, a weak base, as it was determined by the system. Each of the 12 wells of a 96-well microtitre plate row was pre-set to a specified value (every 0.5 pH from 4.0 to 9.5) and the solubility at each pH point was determined using the aqueous dilution method. The straight line indicates the maximum solubility that can determined for this compound. It is based on the amount of compound used. The values around this line indicate that all the compound has been completely dissolved and that solubility can only be said to be higher than this limit value at that particular pH. For clarity, only the solid points of the graph contain true solubility information.

150 CHLORPROMAZINE

100

50

0 3

4

5

6

7

8

9

10

pH The output for the complete set of standard compounds from Table 1 is shown in Fig. 4. The cosolvent-method was used.

Fig. 4 Solubility-pH profiles for a standard set of compounds. Profiles with a single negative slope are bases and those with a single positive slope are acids. Multiprotic acids or bases are hard to detect from the curves. Piroxicam is an ampholyte with pKas at 5.07 and 2.33. The effects of the lower pKa is outside of the pH range measured. Flurbiprofen has a solubility higher than the chosen limit and is dissolved at all points. All that can be said is that its solubility is >60 µg/mL. Griseofulvin is a neutral compound and as such, does not show any pH dependency. Indomethacin has only one pKa at 4.42 and shows the characteristic curve for an acid. The drop-off above pH 8.5 is reproducible, but so far unexplained.

CONCLUSIONS

-4

7

aqS0app (µg/mL )k

The PSR4s collects 50 to 100 data points for each UV spectrum used in the calculation of solubility. A typical example of the scanned sample and reference spectra for miconazole, a weak base. As precipitation takes place to varying degrees at different pH values, the spectra of the sample solutions change in optical densities, according to Beer’s law.

Fig. 2 Clipped spectra of OD vs. wavelength for sample and reference solutions in the PSR4s. The reference spectra shows hardly any dependence on pH whereas the sample spectra shows that OD has a high value at pH 3.0 systematically dropping to a very low value at pH > 6.5. Miconazole is a weak base with a pKa of 6.07. At pH >> pKa the base is not protonated and therefore has its lowest solubility.

-3

6

pKa

Where ∆ = pKa − pKaapp. For acids: ∆ ≤ 0 and for bases: ∆ ≥ 0.

True solubility-pH profile

-6

Compound

Where CRCOS is the calculated concentration of the sample in the cosolvent reference solution and RCOS = AUC RCOS/AUCRCOS.

CHLORPROMAZINE

-2

Table 1 Comparison of solubility results determined by different technologies. The highlighted compounds serve well as standards to include with a set of unknowns for QC

Solubility (mg/mL)

Demonstrate a new way of extracting aqueous intrinsic solubilities of drug molecules from data distorted by DMSO-drug binding or self aggregation reactions.

[1]

Avdeef, A., 2001. High-Throughput Measurements of Solubility Profiles, In Pharmacokinetic Optimization in Drug Research,

[2]

Avdeef, A.; Berger, C.M.; Brownell, C., 2000. pH-Metric Solubility. 2. Correlation between the Acid-Base Titration and the Saturation Shake-Flask Solubility-pH Methods. Pharm. Res., 17, 85-89.

From the data presented it can be concluded that the high-throughput methods presented produce intrinsic solubility figures that compare very well with those of the classical saturation shake-flask method. A major problem in analyzing stock compounds is their DMSO storage medium. Even small concentrations of 1% v/v or less may cause falsely high solubilities to be reported. For ionizable compounds it is possible to compensate for this problem. At least two data points are needed, one on either side of the pKa of the compound. The same compensation mechanism also compensates for the effects of selfaggregation. For increased throughput in a screening setting, measurements at pH 3 and pH 10 for each compound may be all that is needed. Acids and bases can be identified from the two values if they do not both exceed the limit. Ampholytes with pKas in the pH range would not necessarily be properly classified, whereas neutral compounds of low solubility probably would. The limit of detection for the system is about 0.1 µg/mL.