Anal. Chem. 2000, 72, 261-266
Accelerated Articles
Automated High-Throughput Liquid-Liquid Extraction for Initial Purification of Combinatorial Libraries Sean X. Peng,* Charles Henson, Michael J. Strojnowski, Adam Golebiowski, and Sean R. Klopfenstein
Health Care Research Center, The Procter & Gamble Company, 8700 Mason-Montgomery Road, Mason, Ohio 45040
An automated high-throughput liquid-liquid extraction (LLE) methodology has been developed and utilized for the initial purification of the combinatorial library samples containing unreacted amines and other water-soluble byproducts or impurities. Various extraction solvents were evaluated along with different extraction devices. The LLE method was automated using 96-well-format plates and a robotic liquid-handling workstation. In the optimized LLE method, crude combinatorial library samples were dissolved in a water-immiscible organic solvent, butyl acetate, and added to each well in a 96-well-format plate packed with an inert support material coated with hydrochloric acid. Separation occurs based on the partitioning of the compounds between two liquid phases. Product recovery, purity, and amine removal efficiency were determined by HPLC with and without precolumn derivatization. The automated method was successfully applied to the cleanup of some representative combinatorial library samples with greater than 98% amine removal and an average product purity of 90%. The application of the automated high-throughput LLE method should greatly reduce the labor, time, and cost associated with the purification of combinatorial libraries.
Over the past several years, combinatorial chemistry has gained rapid growth in the pharmaceutical industry because of its ability to produce a large number of compounds with a wide range of structural diversity in a short amount of time. Combined with high-throughput screening, computational chemistry, and automation of laboratory procedures, the combinatorial chemistry approach has led to a significantly accelerated drug discovery * Corresponding author: (phone) (513) 622-3944; (fax) (513) 622-3681; (email)
[email protected]. 10.1021/ac990946v CCC: $19.00 Published on Web 12/08/1999
© 2000 American Chemical Society
process compared to a traditional one-compound-at-a-time approach.1 During the high-throughput biological screening of combinatorial compounds, initial sample purification to remove assay-interfering components is required to ensure true hits from various screening assays. Otherwise, false positive response will result. This initial sample cleanup step poses a great challenge to synthetic and analytical chemists since high-throughput and automation will be needed to keep up with the fast pace of combinatorial synthesis and high-throughput screening. In general, combinatorial chemistry employs either solid-phase or solution-phase synthesis.2 The solid-phase synthesis has the advantage of generating cleaner samples since the solid support material, i.e., resin beads, can be filtered and washed after the reaction is complete and purer reaction product is obtained upon cleavage of the linker on the resin. However, excess cleaving reagents are often needed to achieve complete cleavage in a short amount of time because many cleavage reactions are slow with stoichiometrical amounts. For solution-phase synthesis, excess starting reagents and byproducts often remain in the intermediate and final product and are difficult to remove. When those excess reagents and byproducts interfere with biological screening assays of interest, the screening results become unusable. To address those problems, tremendous ongoing efforts have been made in the purification of combinatorial library products. Currently, parallel column HPLC technology has been employed to increase the sample throughput by permitting more than one sample to be analyzed and purified at a time.3 However, the throughput of this methodology is still limited as combinatorial compounds are generated rapidly with the use of 96- or 384-well plates. This parallel method is best suited for purification of hit samples for hit identification and confirmation as the number of samples to (1) Dolle, R. E.; Nelson, K. H. J. Comb. Chem. 1999, 1, 235-282. (2) Czarnik, A. W. Anal. Chem. 1998, 70, 378A-386A. (3) Zeng, L.; Kassel, D. B. Anal. Chem. 1998, 70, 4380-4388.
Analytical Chemistry, Vol. 72, No. 2, January 15, 2000 261
Figure 1. Structures of amines and final combinatorial library products.
be purified is significantly reduced after initial biological screening. For the initial biological screening, however, the known screening assay-interfering components should be removed to avoid false positive responses. To keep pace with the high-throughput combinatorial synthesis, the high-throughput initial sample cleanup is critical for biological screening, which is followed by parallel HPLC purification of hit samples. This initial sample cleanup or purification approach is attractive and cost- and time-effective when it can be automated in a 96-well format. Amines have been used extensively as starting or cleaving reagents for the generation of combinatorial libraries in both the solution- and solid-phase syntheses. However, amines have also been known to interfere with many biological assays, causing false positive results from the screening. Therefore, it is necessary that the excess amines be removed from the final products prior to the biological screening. Since combinatorial library compounds are generally produced in a 96- or 384-well-format, an automated high-throughput sample cleanup procedure is required to avoid the bottleneck in the drug discovery process. Solid-phase extraction (SPE) is a well-established technique for extracting the desired compounds from complex matrixes. The extraction can be fully automated with the use of 96-well SPE plates and a programmable 96-channel liquid-handling workstation.4,5 However, this technique is not effective when the compounds to be extracted, such as basic drugs, are adsorptive to the surfaces of sorbent materials. In addition, SPE methods are generally sensitive to the compounds to be extracted. The liquid-liquid extraction (LLE) is well-established and the most widely used extraction technique in organic synthesis. It is simple, rapid, and convenient. LLE has the advantage of producing extremely clean extracts and (4) Janiszewski, J.; Schneider, R. P.; Hoffmaster, K.; Swyden, M.; Wells, D.; Fouda, H. Rapid Commun. Mass. Spectrom. 1997, 11, 1033-1037. (5) Rossi, D. T. LC-GC 1999, 17, S4-S8.
262
Analytical Chemistry, Vol. 72, No. 2, January 15, 2000
high recovery of the extracted analytes compared to SPE. The main drawback of LLE is that the procedure is often not amenable to automation. Recently, LLE methodology has been employed and reported for the sample cleanup of combinatorial libraries along with solid-supported scavenger methodology to selectively remove unreacted excess starting materials.6-9 Here we present an automated high-throughput LLE method using 96-well-format plates and a robotic liquid-handling workstation for the removal of amines and other water-soluble byproducts and impurities from crude combinatorial samples. Optimal conditions are reported based on results from several representative amines, extraction solvents, and devices. The LLE method developed is fully automated, high-throughput, and efficient for the removal of amines and other water-soluble impurities and byproducts in the combinatorial library samples. EXPERIMENTAL SECTION Materials. Amines used in the combinatorial synthesis (shown in Figure 1), 4-methylbenzylamine (I), 4-methoxybenzylamine (II), 4-fluorobenzylamine (III), heptylamine (IV), cyclohexylamine (V), 5-amino-1-pentanol (VI), piperonylamine (VII), 2,2-diphenylethylamine (VIII), and naphthylethylamine (IX) were purchased from Aldrich (Milwaukee, WI). Combinatorial library products were obtained from Procter & Gamble Pharmaceuticals (Mason, OH). Hydrocholoric acid, chloroform, dichloromethane, 1-butanol, 1-hexanol, ethyl acetate, butyl acetate, and phenyl isothiocyanate (PITC) were obtained from Aldrich (Milwaukee, WI). HPLC-grade (6) Cheng, S.; Comer, D. D.; Williams, J. P.; Myers, P. L.; Boger, D. L. J. Am. Chem. Soc. 1996, 118, 2567-2573. (7) Johnson, C. R.; Zhang, B.; Fantauzzi, P.; Hocker, M.; Yager, K. M. Tetrahedron 1998, 54, 4097-4106. (8) Kaldor, S. W.; Siegel, M. G.; Fritz, J. E.; Dressman, B. A.; Hahn, P. J. Tetrahedron Lett. 1996, 37, 7193-7196. (9) Booth, R. J.; Hodges, J. C. J. Am. Chem. Soc. 1997, 119, 4882-4886.
acetonitrile, formic acid, phosphoric acid, and monobasic sodium phosphate were purchased from J. T. Baker (Phillipsburg, NJ). Diatomaceous earth material, 96-well plates with frits, and 96-well collection plates were purchased from Varian (Harbor City, CA). The 96-well hydrophobic GF/C glass fiber filter plates were obtained from Whatman (Clifton, NJ). Preparation of Extraction Solvents and Samples. Six organic extraction solventsschloroform, dichloromethane, 1-butanol, 1-hexanol, ethyl acetate, and butyl acetateswere evaluated. Each amine test sample was prepared at 10 mg/mL in the respective organic extraction solvents in the 96-well plates. Two types of aqueous extraction solutions, hydrochloric acid and a phosphate buffer, were compared. Hydrochloric acid solutions were prepared at 1, 2, and 4 N by diluting appropriate amounts of the concentrated HCl with Milli-Q water. A 1.0 M phosphate buffer at pH 2.0 was made by dissolving an appropriate amount of monobasic sodium phosphate in Milli-Q water, and the pH was adjusted by adding an appropriate amount of phosphoric acid. The crude combinatorial library samples in the 96-well plate were made by adding 800 µL of the selected organic extraction solvent into each well containing crude dry products. Extraction Apparatus and Procedures. A 96-channel programmable liquid-handling workstation (Quadra 96, model 320, Tomtec, Hamden, CT) was utilized for automation of the LLE procedures for test amines and combinatorial library samples. Three different types of 96-well extraction plates were evaluated: 96-well plate with hydrophobic GF/C glass fiber bottom filter, diatomaceous earth filled 96-well plate with polyethylene bottom frit, and diatomaceous earth packed 96-well plate with hydrophobic GF/C glass fiber bottom filter. When diatomaceous earth packed plates ( ∼200 mg/well) were used, 800 µL of hydrochloric acid was first loaded to the plate and 800 µL of amine samples dissolved in the respective organic solvent was then added to the plate. Subsequently, 800 µL of organic extraction solvent was used to elute the extracted samples into the collection plate, during which a low vacuum (3 in.Hg) was applied to aid in the elution. When a 96-well plate with hydrophobic bottom filter was evaluated, 800 µL of hydrochloric acid and 800 µL of amine samples dissolved in the organic extraction solvent were sequentially loaded to the plate. The organic layer was collected after six cycles of mixing of the aqueous hydrochloric acid and chloroform layers. All those steps were automated by the Quadra 96 liquid-handling workstation. All library samples were prepared and processed in the same manner as that described for the amine samples. Chromatographic Equipment and Conditions. HPLC was performed on a Waters (Milford, MA) Alliance HPLC/PDA system consisting of a 2690 separations module and a 996 photodiode array detector. The system was controlled by the Waters Millennium 2020 data system. A Waters Symmetry C18 column (150 × 3.9 mm i.d., 5 µm) was utilized. A 20-min linear gradient elution from mobile phase A (acetonitrile/water/formic acid; 5:95:0.1; v/v/v) to mobile phase B (acetonitrile/water/formic acid; 95:5: 0.1; v/v/v) was employed for all samples. The flow rate was 1.0 mL/min with UV detection at 260 nm for all samples. For the aromatic amines that contain strong chromophores, samples (10 µL) before and after the LLE were directly injected onto the HPLC system. For the alkylamines that lack strong chromophores, 5 µL of PITC reagent was added to 100 µL of the
Table 1. Percent Amine Removal after Automated High-Throughput LLE Using Extraction Solvents, Chloroform, and Hydrochloric Acid, in Three Different Extraction Devicesa diatomaceous earth amine
without hydrophobic filter
with hydrophobic filter
hydrophobic filter onlyb
% removal I VIII IX
99.2 (0.1) 98.1 (0.3) 98.5 (0.2)
99.4 (0.2) 98.7 (0.2) 99.0 (0.4)
86.3 (0.4) 84.7 (0.6) 81.2 (0.5)
a Values are determined by HPLC and expressed as mean with % RSD in parentheses, n ) 5. b Without diatomaceous earth.
collected LLE eluate and samples before the LLE to form UVabsorbing phenylthiocarbamoyl derivatives (reaction at room temperature for 10 min). Subsequently, 10 µL of the resulting solution was injected onto the HPLC system. RESULTS AND DISCUSSION Selection of Extraction Devices. Three different types of extraction devices were evaluated for the efficiency of amine removal and convenience of extraction and automation. The first extraction device was the diatomaceous earth filled 96-well plate with polyethylene bottom frit. The second extraction device was similar to the first one except that the polyethylene bottom frit was replaced with the hydrophobic GF/C glass fiber filter. The third type of extraction device was a 96-well hydrophobic GF/C glass fiber filter plate without diatomaceous earth material. During the evaluation of these extraction devices, chloroform and 2 N hydrochloric acid were used as the organic and aqueous extraction solvents, respectively. When the extraction was conducted with the first two devices, hydrochloric acid, amine samples, and chloroform were sequentially added to each plate using the Quadra 96 liquid-handling workstation. The eluate in the collection plate was obtained for HPLC analysis. When the third extraction device was used, equal volumes (800 µL) of hydrochloric acid and amine samples dissolved in chloroform were added to the plate and mixed well by the Quadra 96. The bottom chloroform layer was collected for HPLC analysis. Table 1 lists the extraction results using these three types of extraction plates. As shown in Table 1, optimal percent removal of three different amines was achieved using the 96-well plate packed with diatomaceous earth material and sealed by the bottom hydrophobic filter. Clearly, the high surface area of the diatomaceous earth packing material provided continuous and efficient extraction of amines between the two liquid phases. The bottom hydrophobic filter was also useful in that it helped to retain the hydrochloric acid solution containing extracted amines. On the basis of the above results, the 96-well plate with bottom hydrophobic filter was chosen as the extraction device of choice for the removal of amines and water-soluble impurities in our subsequent studies. Selection of Extraction Solvents. The selection of extraction solvents for LLE is based on many factors, depending on the purpose of the extraction. For our applications, the selection of the aqueous extraction solvent was relatively straightforward. Hydrochloric acid and a pH 2 phosphate buffer were chosen as the aqueous extraction phases to be evaluated because the basic Analytical Chemistry, Vol. 72, No. 2, January 15, 2000
263
Figure 2. Four pairs of representative chromatograms of amines, I (A), III (B), V (C), and IX (D), dissolved in butyl acetate before (upper trace) and after (lower trace) the automated LLE.
compounds, i.e., amines, were to be extracted from the organic phase into the aqueous phase. For the selection of the waterimmiscible organic solvent, however, many factors should be considered. Among them, polarity index, viscosity, density, water solubility, and safety profiles (toxicity and flammability) are commonly compared. A good organic extraction solvent should possess a high partition coefficient (distribution ratio) for the solute of interest, low solubility in the aqueous phase, and low toxicity and flammability. On the basis of the above consideration, the following six organic solvents were selected and tested: chloroform, dichloromethane, 1-butanol, 1-hexanol, ethyl acetate, and butyl acetate. When a phosphate buffer (1.0 M at pH 2.0) and a series of hydrochloric acid solutions (1-10 N) were evaluated for the removal of 10 mg/mL amines, the results indicated that the removal of amines was more than 10% better from hydrochloric acid than from the phosphate buffer. Furthermore, 2 N hydrochloric acid showed the same amine removal efficiency as that from 10 N hydrochloric acid. Therefore, 2 N hydrochloric acid was chosen as the aqueous extraction phase. Among the organic solvents tested, chloroform and butyl acetate gave the best results in terms of percent amine removal. As shown in Table 2, the removal of eight representative amines is better than 95% when chloroform and butyl acetate were used. Furthermore, the amine removal is consistently higher in butyl acetate than in chloroform. The additional fact that chloroform is highly 264
Analytical Chemistry, Vol. 72, No. 2, January 15, 2000
toxic and tends to erode the 96-well plate material (polypropylene) has led us to select butyl acetate as the final organic extraction solvent for amine and library samples. Automation. Automation of the LLE method was accomplished using a 96-channel liquid-handling workstation (Quadra 96). The Quadra 96 was programmed to perform the LLE as follows. The Quadra 96 prepared the sample plate by aspirating 800 µL of organic extraction solvent from a reservoir and loading the solvent into the 96-well sample plate containing amines or combinatorial samples. The initial dry samples were dissolved and mixed in the organic solvent by six aspirate-dispense mixing cycles by the Quadra 96. Then the Quadra 96 conditioned the extraction plate by sequentially aspirating hydrochloric acid from a solvent reservoir and dispensing it to the extraction plate. A hydrochloric acid layer was thus formed on the surface of the diatomaceous earth material in each well. The LLE extraction occurred when the Quadra 96 aspirated 800 µL of amine samples from the sample plate and loaded it into the extraction plate. After ∼5 min of extraction, the Quadra 96 picked up 800 µL of the organic extraction solvent and added it into the extraction plate to elute any remaining extracted organic layers into the collection plate (to improve product recovery). All the above steps were fully automated and carried out by the Quadra 96. The elution step was conducted using a vacuum manifold under a low vacuum (∼3 in.Hg) to shorten the elution time. The whole extraction procedure
Table 2. Amine Removal, Product Recovery, and Purity of a Partial Combinatorial Library before and after Automated LLE Using a Diatomaceous Earth 96-Well Plate with a Bottom Hydrophobic Filtera amine I
II
III
IV
V
VI
VII
VIII
99.4 (0.2)
Amine Only with Elution Solvent: Chloroform 95.7 (0.6) 98.6 (0.5) 98.1 (0.2) 95.9 (0.3)
98.3 (0.4)
98.7 (0.3)
98.7 (0.2)
% amine removal
99.9 (0.1)
Amine Only with Elution Solvent: Butyl Acetate 98.9 (0.2) 99.2 (0.2) 98.5 (0.1) 99.3 (0.6)
99.9 (0.1)
99.1 (0.4)
98.9 (0.1)
% amine remaining before LLE % amine remaining after LLE % final product recovery % product purity before LLE % product purity after LLE
Amine and Product with Elution Solvent: Butyl Acetate 15.5 (0.9) 8.2 (0.5) 12.3 (0.8) 11.4 (0.8) 18.6 (0.7) 0.02 (0.9) 0.09 (0.5) 0.10 (0.8) 0.17 (0.8) 0.12 (0.7) 85.8 (1.1) 97.3 (0.6) 88.1 (0.9) 86.0 (1.1) 91.3 (0.5) 83.2 (0.6) 73.8 (0.9) 76.4 (1.2) 79.6 (1.0) 72.9 (0.8) 91.6 (0.4) 90.1 (0.7) 85.6 (1.0) 88.3 (0.9) 90.5 (0.6)
18.1 (0.9) 0.13 (0.8) 88.3 (1.0) 80.4 (0.7) 92.1 (0.9)
22.6 (0.6) 0.25 (0.8) 94.5 (0.7) 70.2 (1.1) 90.8 (0.7)
15.2 (1.0) 0.21 (1.2) 91.9 (0.9) 81.5 (0.8) 91.3 (1.1)
% amine removal
a
Values are determined by HPLC and expressed as mean with % RSD in parentheses, n ) 5.
took ∼15 min. HPLC Assays. A generic reversed-phase HPLC method was developed and employed to determine the efficiency of amine removal and recovery of the final combinatorial products. Amine and library samples before and after LLE were injected onto the HPLC system; the amine and product peak areas were then compared to assess the amine removal and product recovery. The alkylamines that do not contain strong UV chromophores were subjected to derivatization with PITC before and after the LLE prior to HPLC analysis. The representative chromatograms from the HPLC assays are shown in Figure 2. As can be seen from Figure 2, the amines or derivatized amines are well separated from other components. No interference peaks were observed. Figure 3 shows representative chromatograms from two combinatorial samples before and after LLE. The amine and product peaks are also well resolved and separated from other component peaks. The generic HPLC method employed for all samples was simple and reproducible. The precision expressed as the relative standard deviation based on six repetitive injections was less than 5% and the accuracy was in the range of 95-103% at amine concentrations of 10 and 500 µg/mL in chloroform or butyl acetate. Extraction Selectivity and Efficiency. The extraction selectivity and efficiency are determined by amine removal and product recovery from the crude combinatorial samples. In addition to amines, other water-soluble components are also removed during the extraction to a certain extent depending on their partition coefficients between the two extraction phases. These are evidenced in the chromatograms where some component peaks disappear from the extracted eluate. The removal of the amines is more than 98% while the recovery of the final products is generally greater than 85%, for the amine and combinatorial samples tested, depending on the partition coefficients of the amine and product. In this automated LLE method, the total time required to complete the extraction of one 96-well plate of crude library samples is generally less than 15 min. Therefore, four 96well plates of combinatorial samples can be extracted per liquid handler per hour. Applications. This automated LLE methodology was applied to the initial purification of combinatorial library samples in a 96well format. The removal of amines was efficient as shown in Table 2. Before LLE, the crude partial library samples contained excess amines in the range of 8.2s22.6%. After LLE, less than 0.25% of
Figure 3. Two pairs of representative chromatograms of combinatorial library samples before (upper trace) and after (lower trace) the automated LLE.
the amines remained in the samples. The percent recovery of the final combinatorial product ranged from 85.8 to 97.3%. The variations in product recovery were most likely due to the variations in partitioning properties of the structurally different Analytical Chemistry, Vol. 72, No. 2, January 15, 2000
265
products studied. In addition to amines, other water-soluble byproducts or impurities were also removed during the LLE extraction, resulting in final products of higher purity. As shown in Table 2, the purity of the final product improved significantly after the LLE to greater than 85%. This automated high-throughput LLE method is generally applicable and can be applied to the removal of other basic and neutral water-soluble components in combinatorial library samples. In the case where most of the components to be removed are acidic, basic buffer solutions may be used instead of hydrochloric acid as an aqueous extraction phase. CONCLUSIONS An automated high-throughput liquid-liquid extraction methodology has been developed and utilized for the removal of excess amines and water-soluble impurities from combinatorial library samples. The method is simple, rapid, and reliable. After evaluation of different extraction solvents and devices, optimal extraction conditions were obtained by employing butyl acetate as an organic solvent and hydrochloric acid as an aqueous solvent in a diato-
266
Analytical Chemistry, Vol. 72, No. 2, January 15, 2000
maceous earth packed 96-well plate with bottom hydrophobic filter. This automated method was applied to the initial purification of crude combinatorial library samples in a 96-well format. Among nine amines tested, greater than 98% removal of excess amines was achieved and more than 85% recovery and better than 85% purity (average purity of 90%) were obtained from the partial combinatorial library samples. The methodology can be modified to remove acidic and neutral water-soluble components by selecting appropriate aqueous extraction solvents. In conjunction with parallel preparative HPLC techniques for further hit purification and identification, this automated LLE methodology should provide great benefits to initial purification of crude library samples for high-throughput screening, which would ultimately result in reduced labor, time, and cost associated with the purification of combinatorial libraries.
Received for review August 19, 1999. Accepted November 11, 1999. AC990946V
