ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 OH 39 Short Papers in Pharmaceutical Analysis The following are summaries of five of the papers presented at a Meeting of the Joint Pharmaceutical Analysis Group held on' October 17th, 1991, in the Royal Pharmaceutical Society of Great Britain. Development and Application of an Assay Procedure for Hyoscine in Urine Using Solid-phase Extraction and High-performance Liquid Chromatography with Electrochemical Detection Peter R. Hurst British Pharmacopoeia Commission Secretariat, Market Towers, I Nine Elms Lane, London SW8 5NQ Hyoscine (scopolamine) is a potent anti-cholinergic alkaloid (Fig. 1). It is used as a pre-medicant and also to counteract motion sickness, for which purpose it may be administered orally or via a transdermal patch.A transdermal patch delivers an average of 500 pg of hyoscine over 72 h, producing an estimated steady-state plasma concentration of around 100 pg ml-'.' The compound is unstable; it decomposes to apohyoscine on gas chromatography columns and at high pH values it hydrolyses to scopoline and tropic acid (Fig. 1). This instability together with the low concentrations expected poses severe analytical problems. Analytical approaches to date include a gas chromatography-mass s ectrometry (GC-MS) method' and radioreceptor assays.'-' A high-performance liquid chromatographic (HPLC) method with electrochemical detection, suitable for the measurement of hyoscine in urine after oral doses o r administration of a transdermal patch, is presented in this paper. Experimental Transdermal patches (Scopoderm TTS, Ciba Laboratories, Horsham, Sussex) were applied to volunteers on two occasions one week apart on a double blind cross-over basis.Urine samples collected were assayed before and after incubation with P-glucuronidase-aryl sulfatase solution (Boehringer, Lewes, Sussex). Chromatography Compounds were chromatographed using a straight-phase Spherisorb S5W silica column 150 x 4.6mm (Phase Sepa- Scopoline H<p--> 0 I co I Hyoscine H C-CH 2 0 H O-H 0 I co I C=C H 2 Apohyoscine Fig. 1 Decomposition of hyoscine40 75 50 25 a m 3 0 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 (a) (b) - - 80 Yo 50 % - - - - Lc -L rations, Queensferry, Clwyd) with an eluent consisting of acetonitrile, methanol and 0.1 mol I-' ammonium nitrate buffer pH 9.0 in the proportions (7 + 2 + 1 v/v/v).Detection was via an Environmental Science Associates Model 5100A Coulochem detector. The compounds were oxidized at +0.9 V. & 75 a 50 25 0 Solid-phase Extraction Method Development Tritiated hyoscine and apohyoscine were employed to deter- mine the most suitable solid-phase extraction columns for the assay and also to develop a method capable of specifically extracting hyoscine but not apohyoscine. Urine samples spiked with the tritiated compounds were extracted on to a variety of solid-phase extraction columns and then eluted from them using successive 100 pl aliquots of eluent. Measurement of the radioactivity in each eluted fraction allowed an elution 'profile' to be produced for each sorbent type, allowing subtle yet significant differences in recovery and elution characteristics to be revealed. This approach has also been successfully applied to the development of an assay procedure for physostigmine.' Two-column Solid-phase Extraction Method Urine samples ( 5 ml pre-incubation and 1 ml post-incubation) were adjusted to pH 9 by addition of an equivalent volume of 0.1 mol 1-' borax.Internal standard (250 ng N-ethyl norhyo- scine in 50 p1 of water) was added and samples centrifuged. The supernatants were extracted via alkali-prepared large reservoir 100 mg CI8 columns. The columns were washed with 0.1 mol I-' borax (3 ml), water (3 ml) and 20% methanol- water (1 ml) before eluting the compounds with 1 ml of 50% methanol-water directly into 9 ml of water which had been placed in the reservoirs of acid-prepared 100 mg CN columns.The water-eluate in each reservoir was mixed, extracted via the CN column and the column washed with water (3 ml). The compounds were eluted with 500 pl of HPLC eluent, 200 pl of which was transferred to autosampler vials and centrifuged before injecting on to the chromatogram. (d) (el ( f ) - - - 80% 50 % 40 % c - - - - - 1 1 1 1 1 1 l ~ - Results and Discussion Using the chromatographic system above the resolution of the internal standard (N-ethyl norhyoscine) and apohyoscine was incomplete. However, the possible interference of apohyos- cine was avoided at the extraction stage by judicious choice of elution solvent. The use of 1 ml of 50% methanol-water to elute the C18 column meant that almost all of the hyoscine was eluted whilst apohyoscine remained on the column (Fig.2). The use of 2 x 1 ml of 20% methanol-water allowed many of the urine pigments to be washed from the column without any loss of the compounds of interest. However, some components of the extracted urine sample remained on the column, were co-eluted in the eluate and subsequently interfered with the 'binding' of the compounds of interest with the CN column. Consequently, it was essential that the eluate eluted from the CI8 column was diluted with water before extraction. For this assay diluting the eluate to 5% of its original concentration was sufficient to ensure that all the compounds of interest were extracted. As a consequence of the large number of sample manipula- tions this method has a lower recovery (64%) than alternative approaches (for example a toluene extraction followed by solid-phase extraction gives a recovery of about 85%).This was not considered a problem as the assay was sufficiently sensitive (lower limit of quantitation 2 ng mI-'), precise (coefficients of variation of 4.7, 4.1 and 11 at 100, 10 and 2 ngml-', respectively), linear over the range 1-100 ng ml-' (Y = 0.9998) and gave much cleaner chromatographic traces than any produced by techniques involving liquid-liquid extraction steps. Indeed it is worth noting that with regard to the assay of Table 1 Concentrations of hyoscine in the urine of volunteers after administration of a transdermal patch Subject 1 2 3 4 5 6 7 8 9 10 11 12 Hyoscine concentrationhg ml-' ~ Total? Free* 7 122 13 55 8 67 5 35 12 121 7 72 4 31 9 69 19 128 9 99 19 118 9 94 * Free = concentration before incubation with (3-glucuronidase-aryl t Total = concentration after incubation with P-glucuronidase-aryl sulfatase at 37 "C for 24 h.sulfatase at 37 "C for 24 h. 10 20 20 10 20 Fraction number (0.1 rnl) Fig. 2 extraction of each compound from 5 rnl urine on to the alkali prepared C18 column) Elution of hyoscine [(a)-(c)] and apohyoscine [ ( d ) - ( f ) ] from a 100 mg C18 column using varying percentages of methanol-water (afterANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 41 drugs from biological matrices (e.g., blood, plasma, urine) a References sample preparation step which results in a decreased recovery of the analyte may be accompanied by an increase in sensitivity of the assay.Assay results (Table 1) indicate that a large proportion of the hyoscine present in urine is present in a conjugated form, only about 10% of the total being present as hyoscine. The presence of apohyoscine (which was shown to be present in some urine samples) was significant as this casts some doubt upon the results of workers who assayed hyoscine indirectly by measur- ing its hydrolysis product scopoline.’ Scopoline would also be formed by hydrolysis of apohyoscine (Fig. 1). 1 Walker, G., and Massam, D., ABPl Data Sheet Compendium IYYl-92, Datapharm Publications Ltd., London, p. 313. 2 Bayne. W. F., Tao, F. T., and Crisologo, N., J. Phurm. Sci., 1975, 64, 288. 3 Metcalfe, R. F.. Biochem. Pharmacol.. 1981, 30, 209. 4 Cintron, N.M.. and Chen, Y.-M., J. Phurm. Sci., 1987, 76, 328. 5 Kentala, E., Kaila, T., Ali-Melkkila, T., and Kanto, J., lnt. J . Clin. Pharmacol. Ther. Toxicol., 1990, 28, 487. 6 Hurst, P. R., and Whelpton, R.. Biomed. Chromatogr., 1989.3, 226. Direct Non-destructive Colour Measurement of Pharmaceuticals Nicola Stock Pharmaceutical Analysis Department, Glaxo Group Research Ltd., Ware, Hertfordshire SG 72 ODP This paper briefly discusses colour measurement techniques with emphasis on tristimulus measurements. Recent work to modify a commercially available colour measurement system to accommodate the measurement of new pharmaceutical preparations in a direct, non-destructive manner is described. Colour is a part of our everyday lives and is a phenomenon that is taken for granted.Measurement of colour is undertaken for various reasons. In paint and cosmetics industries, for example, the final colour and consistency of colour of the product are vital, whereas in the pharmaceutical industry colour measurement can provide information about the formation of coloured impurities or can be used to control a manufacturing process. Colour is not an absolute measurement; people have different perceptions of colour and perceived colour is dependent on such parameters as particle size, texture, shape, angle of observation and distance. Colour measurement can be divided into two categories: operator observation and instru- mental techniques. Operator observation can be as simple as a description, t?.g., a yellow liquid, or can be comparative such as in pharmaco- poeial techniques, where a sample can be compared against a range of colour standards (for instance BY, GY solution series found in the British Pharmacopoeia).Instrumental techniques include spectrophometric and colorimetric methods. APHA measurement is an example of a spectrophometric technique where an inorganic colour standard at different strengths is measured at a particular wavelength, a calibration graph constructed and the sample then measured against that graph. Tristimulus measurements are derived from the amount of light transmitted or reflected from a sample in the visible range. Instrumental techniques are to be preferred because they are objective and do not rely on visual comparison of inorganic standards with organic samples. Tristimulus measurements are the most convenient and reliable way of measuring colour because they can be manipulated into readily understandable scales. , Tristimulus Colour The eye can detect three types of colour variation via three receptors: hue, brightness and saturation.The response of the eye has limiting wavelengths at 380 nm (deep violet) and 780 nm (deep red). Beyond these limiting wavelengths the invisible part of the electromagnetic spectrum is reached. Colour can be broken down into thrce radiant stimuli which excite the three receptors in the eye. It is possible to mix the three colour primaries red, green and blue to match light of any given colour. In practice this results in a little of each primary being added to the colour under test to obtain a match.This can be represented thus: Colour = Y(R) + g(G) + b(B) (1) where Y, g and b are the equivalent amounts of red, green and blue primaries. The measurement of these amounts corres- ponds to the tristimulus value. Transcription of these values into co-ordinate form causes practical problems which led the Commission Internationale de I’Eclairage (CIE), the body who are responsible for setting colour standards, to carry out some colour matching experi- ments with volunteers of normal colour vision. The distributing coefficients for each visible wavelength (400-700 nm) giving relative amounts of stimulation for each receptor caused by light of that wavelength were measured. The resulting curves (Fig. 1) (1931 CIE system for a two-degree field of view) give the spectral distribution and amounts of idealized primaries required to match any colour in the visible range.The y curve is the spectral luminous efficiency curve for the photopic eye and therefore gives lightness information, while the x and z curves give colour information only. The amount of each primary required to colour match is the tristimulus value. Tristimulus measurements can be reduced to a two-dimen- sional chromaticity plot (Fig. 2), where x and y are the nor- malizcd tristimulus [x = X/(X+Y+Z), y = Y/(X+Y+Z)]. The 350 42 5 500 575 650 725 Wavelengthlnm Blue Green Red Fig. 1 Spectral distribution coefficients42 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 Y Magenta 0 1 4 0 6 2 3 4 5 6 7 8 X Fig. 2 CIE chromaticity diagram enclosed area contains all the colours which can be detected by the human eye.For a perceived red colour a positive amount of red and a negative amount of blue and green radiation are seen, producing a redder than red imaginary primary. Chroma- ticity mesurements have disadvantages in that they do not represent equal colour differences as equal distances in space. The CIE have created more simple scales, the CIELUX (for television and graphic arts) and the CIELAB (e.g., cosmetics, paint, pharmaceuticals). The LAB values are a mathematical transformation of tristimulus data defined by three axes (Fig. 3). The L axis is the third dimension lightness scale derived from the tristimulus Y by a non-linear transformation, A and B are Cartesian coordinates where a positive A value represents redness and a negative value greenness, a positive B value represents yellow and a negative one blue. The understanding of colour can be further clarified by expressing the values as polar coordinates (Fig.3). The distance of the colour from the origin is the chroma or strength of colour and the actual colour is the hue angle in the range 0- 360" counter-clockwise, where 0" is red, 90" yellow, 180" green, 270" blue, back to red at 360". LAB values can be expressed using the derived value DE: A E = (AL? + AA: + AB?); (2) where A L , A A and AB are the differences between colour coordinates of two samples being compared, e.g., a sample and its recognized standard or a stressed stability sample with the controlled temperature sample. The eye can detect a nE value T White L = 100 Light Fig. 3 LAB and LCH colour coordinates of about 4.This value will demonstrate a difference between samples but will give no information about what that difference is. Tristimulus Instrumentation Instruments commonly available are the Trivector colour measurement systems, the most recent model being the CL6000 which is available in both reflectancc and transmission modes. The pharmaceutical industry uses the reflectance mode for solid colour assessment such as tablets, whereas the trans- mission mode is used for liquids such as injectable prep- arations. Both types of measurements are performed using transmission or reflectance measuring heads which are con- nected to a control unit which can be linked to a personal computer with appropriate software (Fig.4). The transmission head measures the liquid sample placed in either a 1 cm o r 5 cm UV cell, whereas the reflectance head usually measures the sample directly. Experimental The development of a new formulation at Glaxo Group Research involving very small volumes (125 PI) of liquid in unit dose devices presented an interesting problem for conventional instrumentation. The use of a 1 cm cell would involve bulking a large number of sample devices, which is not only wasteful but gives no single device information. Small volume flow cells were also investigated but problems occurred with the width of the light beam in the CL6000 transmission head and more than one sample device was still necessary. Therefore, a direct method of measuring colour was required, preferably in a non- destructive manner so that a single sample device could be used to gain colour information and then be used for other analyses such as assay and impurities, thereby enabling correlation of colour with possible degradation profiles.Development Early work suggested the use of fibre optic cables to pass light directly through a sample. A prototype system was developed (Fig. 5 ) which directed light from an independent light source via fibre optic cables through a sample device into a CL6000 transmission head. Initial sample repositioning problems were overcome by design of a crude sample holder to keep the device rigid, in order to minimize optical effects from movement of the glass sample container. Tables 1 and 2 compare results with and without the sample stand and demonstrate that excellent reproducibility can be obtained with the stand.Some varia- bility was seen between sample devices which can be attributed to the nature of the sample. The success of this prototype system led to the commercial manufacture of a modified Trivector CL6000 transmission DL-,n Optional PC 7 1 cm or 5 cm cells I I Control unit Transmission head Fig. 4 Schcmatic diagram of a conventional Trivector CL6000 transmission systcmANALYTICAL PROCEEDINGS. JANUARY 1993. VOL 30 43 Transmission Sample Fig. 5 guides Schematic diagram of prototype system with fibre optic light Table 1 Tristimulus measurements without sample stand Chromaticity [x = ( X x lOOO)l(X + Y + Z ) ] Tristimulus Position X Y 2 X Y I 381 327 73 488 419 2 285 251 54 483 426 3 280 244 53 485 423 4 372 320 69 488 420 5 296 252 54 492 419 6 219 192 42 483 424 7 244 195 44 484 42 1 8 243 213 46 484 424 9 244 215 49 480 423 10 394 341 72 488 423 Mean 294 255 56 486 422 RSD(%) 66 56 12 3 2 Table 2 Tristimulus measurements with sample stand Chromaticity [x = ( X X 1000)l(X + Y + Z ) ] Tristimulus Position X Y Z X Y I 2 3 4 5 6 7 8 9 10 261 228 46 263 229 46 265 231 47 264 230 47 265 231 47 262 229 46 265 231 47 261 228 46 263 230 47 263 230 47 488 489 488 488 488 488 488 488 487 487 426 426 225 425 425 426 425 426 426 426 Mean 263 230 47 488 426 RSD(Yo) 1 1 1 1 0 Transmission head Sample holder Optional PC Fig.6 Schematic diagram of modified CL6000 transmission system Table 3 Type 1 sample device repositioned ten times using the modified system Chromaticity [x = ( X x 100O)l(X + Y + Z ) ] Tristimulus Position X Y 2 X Y 1 571 534 76 483 452 2 555 519 75 483 452 3 554 518 75 483 452 4 553 518 75 483 452 5 557 520 75 484 45 1 6 553 517 75 483 452 7 556 520 75 483 452 8 560 523 75 484 452 9 555 519 75 483 452 10 556 519 75 483 45 1 Mean 555 521 72 483 452 RSD (Yo) 1.0 0.9 0.4 0.1 0.1 Table 4 Type 2 sample device repositioned ten times using the modified system Chromaticity [x = ( X X 100O)l(X + Y + Z ) ] Tristimulus Position X Y Z X Y 1 956 964 851 345 348 2 954 963 849 345 348 3 957 966 854 345 348 4 960 968 855 345 348 5 960 969 856 345 348 6 833 841 743 345 348 7 954 963 849 345 348 8 959 967 855 345 348 9 958 967 853 345 348 10 945 954 842 345 348 Mean 944 952 840 345 348 RSD (Yo) 4.1 4.1 4.1 - - system (Fig.6). This differs from the prototype system in that there is n o independent light source and the sample holders have been optically designed to minimize effects from the glass sample container. The validation of the instrument (see Tables 3 and 4) showed excellent reproducibility of measure- ment and comparable results, allowing for differences in path- length between samples measured on the modified system and on an independent conventional system. Different sample holders can be interchanged within the system and the modified optics in the transmission head can be returned to the conventional mode, using 1 or 5 cm cells, in approximately 20 min. The modified system is now in routine use within Glaxo Group Research using two types of sample holders for different pharmaceutical preparations and has proven to be a valuable addition to routine colour measure- ment instrumentation. In conclusion this paper demonstrates that a Trivector CL6000 colour measurement system can be modified for use with fibre optic light guides with the advantages of: non- destructive colour assessment; detection of device-to-device variation for pharmaceutical preparations; correlation of colour and assayhmpurity profiles; rapid, clean technique; possible design of other sample holders; possible conversion back to a conventional instrument; and linkage to a personal computer for easy data manipulation and convenient print-outs.The author is grateful to E. Scorer, Glaxo Manufacturing Services, County Durham and to M.Petty, Cosense Ltd., Cambridge. References Throughout the section on Tristimulus Colour reference has been made to the following: 1 2 Trivector CL6000 System Manual. Unired Srares Phul-mucopeia. 1990, XXIl(1061). 1627.44 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 Trace Analysis of Acyclovir Esters David S. Ashton and Andrew Ray Department of Ph ysical Sciences, Wellcom e Research Laboratories, L ang le y Court, Becken ham, Kent BR3 3BS Acyclovir, 9-(2-hydroxyethoxymethyl)guanine, is an anti-viral drug manufactured by The Wellcome Foundation Ltd. It was the first non-toxic drug to be developed for systemic use against herpes roup viral infections such as cold sores and genital herpes. 'The drug is available for parenteral, oral and topical administration using different formulations.Acyclovir has recently been used to treat shingle^^.^ and a New Drug Application (NDA) has been filed in the USA for the use of acyclovir in the treatment of chickenpox. An analytical method has been established to extract and confirm the presence of esters which are route indicative impurities found during the synthesis of acyclovir. These esters can be important in enforcing patent rights. The method involves liquid extraction of the sample followed by high- performance liquid chromatography (HPLC) in which frac- tions are collected and then further analysed by mass spectrometry. The samples analysed are generally from one of three types: bulk acyclovir powder, tablets and creams (or ointments). The composition of the samples may not be known so a universally applicable method must be used.Experimental Extraction To approximately 1 g of bulk acyclovir powder was added 20 ml of methanol-water (2 + 8). The mixture was shaken for 30 min and then centrifuged for 30 min at 2500 r.p.m. The supernatant liquid was removed by Pasteur pipette and blown down to approximately 2 ml under nitrogen. The sample was then filtered through a 0.45 pm Millex HA filter. An extraction blank was prepared prior to the analysis using the procedure described above to ensure that there was no contamination. HPLC HPLC separations were performed on a Waters 600ms System with a Waters WISP 715 Sample Processor. Separation of the 0 Acyclovir 9-(2-hydroxyethoxymethyl)guanine 0 0-Acetylguanine compound 9-(2-acetoxyethoxymethyl)guanine esters was achieved with a Dupont Zorbax-CS (25 x 0.46 cm) column.A Waters 991 Photo Diode Array detector was used for detection. Data collection and reduction was performed with a VG Multichrom chromatography data system. Water was obtained from a Scorah triple distillation system and HPLC grade methanol was obtained from Rathburns. The HPLC conditions were: methanol-water (2 + 8); flow rate, 1 ml min-'; temperature, 30°C. Detection was by UV moni- toring at 254 nm. Injection size was 100 PI. Mass Spectrometry The fraction corresponding to the compound of interest was analysed by high resolution mass spectrometry. Ions with the mass to charge ratio ('rnlz') of the singly charged molecular ion of the compound of interest were detected by means of high resolution mass spectrometry, using a peak matching tech- nique.The mass spectrometer was a Kratos Concept 1s with a maximum resolving power of 80000 and an accuracy of m/z measurement of 0.0006 a.m.u. at 10 000 resolution in the region of 300 a.m.u. The probe was a direct insertion type. A Sun computer system with Kratos Mach-3 software was used to record the series of scans and to provide averaged results in graphical form. Method Extraction and HPLC Standards of each ester were injected, then a blank injection was made (machine blank) to ensure no residual contamina- tion. This was followed by the extraction blank and finally the sample. Fractions were collected for the two blanks and the sample at the appropriate retention times of the standards.The sample injection was repeated six times and fractions collected. The fractions corresponding to the standard peaks were blown down under nitrogen to approximately 300 pl. These fractions were reinjected on to the HPLC using the same conditions as above. Fractions were again collected and analysed by mass spectrometry. 0 CO I CH3 N-Acetylguanine compound Nz-acetyl-9-(2-hydroxyethoxymethyl)guanine 0 Diacetylguanine compound 9-(2-acetoxyethoxymethyl)-W-acetylguanineANALYTICAL PROCEEDINGS, JANUARY 1993. VOL 30 90 80 45 Peak match + El Scans 87-121 100% = 147 mV - - Sufficient tablets were used to give 1 g of acyclovir. The tablets were crushed using a pestle and mortar and the sample washed into a scintillation vial using 20 ml of methanol-water (2 + 8); the vial was shaken and centrifuged as described previously.The sample was then blown down under nitrogen to approximately 2 ml. For cream or ointment samples a method that removed detergents which might affect the HPLC and mass spec- trometry of the sample was necessary. The method involved dispersing 1 g of cream thoroughly in 20 ml of methanol-water (2 + 8). The sample was shaken for 30 min and centrifuged for 30 min. The centrifuging produced three separate layers, the depth of each being dependent on the composition of the cream. The top layer consisted of any detergents present in the formulation which had flocculated together. This layer was normally a milky white froth. The middle layer was the aqueous phase and was the largest layer, generally a cloudy solution.The third layer was a precipitate on the bottom of the vial containing any heavy organics used in the formulation which were not soluble in the aqueous phase. The middle aqueous layer contained the acyclovir and its related esters. This was carefully removed by pipette to ensure that the other two phases were left behind. This layer was then blown down under nitrogen to 2 ml, filtered through a 0.45 pm filter and injected on to the HPLC as described previously. Mass Spectrometry The mass spectrometer was tuned to an indicated resolution of 10 000 and to its maximum sensitivity for that resolution using an ion of perfluorokerosene with an mlz value of 268.9824 as the reference peak. The instrument was then set to peak matching mode and calibrated against another ion of perfluor- okerosene with an rnlz value of 266.9856.The molecular ion of the O-acetylguanine compound at rnlz 267.0968 was then monitored. A capillary tube was filled with part of the appropriate fraction from the extraction blank and placed in the probe tube. The solvent was removed under vacuum. The probe was then inserted into the mass spectrometer through the vacuum lock and heated gently to a temperature of up to 300°C. The procedure was then repeated with the appropriate fraction collected from the HPLC of the extracted sample. The data acquired was processed by the computer and a graphical output displayed and recorded for each fraction. Results The relatively large amount of acyclovir injected can swamp trace amounts of ester, particularly the N-acetyl ester.It was often necessary to collect fractions about the retention times of the esters, then reduce the volume of the fraction under nitrogen before reinjecting. The re-injected fraction produced a much cleaner chromatogram with more consistent retention times. Fractions from the re-injected run were collected and further analysed by mass spectrometry. This HPLC method was applied to all three esters, whether looking for all three or just one. The only time that the method was changed was when the fraction containing the diacetylguanine compound was re- injected; here the methanol percentage of the mobile phase was increased to shorten the run time and to lower the peak width. A typical analysis of acyclovir cream which attempted to confirm the presence of the O-acetylguanine compound produced Fig.1. The fractions collected on this run and five additional runs were concentrated under nitrogen and re- injected. This fraction, corresponding to the O-acetylguanine compound, was run with a mobile phase of methanol-water (2 + 8) and produced Fig. 2. The fraction from the re-injected run was blown down for analysis by mass spectrometry. Fig. 3 shows the extraction blank monitored for the molecular ion of the O-acetylguanine compound at mlz 267.0968 and Fig. 4 shows the molecular ion of the O-acetylguanine compound at rnlz 267.0968 for the fraction collected from the re-injected HPLC fraction. Discussion Several methods for the analysis of acyclovir have been reported.- These methods are not applicable here due to the instability of the ‘acetyl esters’ at acid pH.The esters undergo a predictable hydrolysis in solution, slowly breaking down to give acyclovir (the diacetyl ester also produces both the mono- acetyl esters). The method described is difficult to quantitate. The overall extraction efficiency is approximately 60%. HPLC can detect the appropriate ester at approximately 0.005 ppm, but this can often be obscured by other components from the preparation. High resolution mass spectrometry detects the accurate mass of the molecular ion of the appropriate ester, demonstrating unequivocally the presence of the ester. Although more sensitive detection can be achieved by both flow liquid secondary ion mass spectroscopy (LSIMS) techniques and thermospray mass spectrometry, these methods do not give the I 1000 1 900 > 800 & 700 $600 500 E 400 300 200 100 - o c a I I I I I 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 Ti me/m i n Fig.1 Extract from a cream m 0 500 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 Time/m i n Fig. 2 Reinjected O-acetylguanine fraction I 0 10 20 30 40 50 60 70 80 90 100 110 120 Channel Fig. 3 fraction rnlz 267.0968 Mass spectrum of extraction blank of O-acetylguanine HPLC46 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 100 90 80 70 60 E $ 50 40 30 20 10 0 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 Channel Fig. 4 from extract mlz 267.0968 Mass spectrum of O-acetylguanine HPLC fraction collected confirmatory accuracy of a high resolution mass spectrometric measurement. References Schaeffer, H.J . , Beauchamp, L . , de Miranda, P., Elion, G. B., Bauer. D. J . , and Collins, P . , Nature, 1978, 272, 583. McKendrick, M. W., McGill, J . I., White, J. E., and Wood, M. J., Br. Med. J . , 1986, 293, 1529. Huff, J . C., Bean, B., Balfour, H. H., Laskin, 0. L., Conner, J . D.. Corey, L., Bryson, Y. J . , and McGuirt, P., Am. J. Med., 1988, 85 (Suppl 2A), 84. Molokhia, A. M., Niazy, E. M., El-Hoofy. S. A., and El- Dardari, M. E.. J . Liq. Chromatogr., 1985, 13, 981. Smith. R. L . , and Walker, D., J . Chromutogr.. 1985, 343, 203. Bouquet. J . , Regnier, B., Quehen, S . . Brisson, A. M., Courtois, P., and Fourtillan, J. B. N., J . Liq. Chromatoer., 1985, 8 , 1663. Investigation into the Reversed Phase High-performance Liquid Chromatographic Behaviour of Tipredane Metabolites and Related Substances Melvin R.Euerby, Christopher M. Johnson* and Steven C. Nichols Analytical Chemistry, Research and Development Laboratories, Fisons Pharmaceuticals plc, Bakewell Road, Loughborough, Leicestershire LEI 1 ORH Tipredane (INN, I), a novel corticosteroid, undergoes oxida- tive metabolism in man to yield the 6~-hydroxy-17-methylsulfi- nyl diastereoisomers (FPL 66365XX and FPL 66364XX) and the corresponding sulfone (FPL 66366XX) derivatives. Syn- thesis of gram quantities of these compounds was required, therefore it was essential that a high-performance liquid chromatographic (HPLC) purity screen was developed which separated the sulfoxide diastereoisomers (FPL 66365XX and FPL 66364XX) from each other and from FPL 66366XX and the related synthesis impurities (FPL 67526XX, and com- pounds I1 and 111).This paper describes the complex chromatographic behav- iour of the tipredane metabolites and related substances, and the subsequent development, using a systematic modular approach, of a selective multisolvent gradient elution HPLC method for the separation of the desired compounds, where solvent strength, selectivity and composition were simulta- neously varied during the analysis. This technique has the advantage that it does not require the use of complex optimization software and can be performed using a binary HPLC delivery system. With this technique the effect of changes in various conditions on the separation of critical pairs of compounds of similar polarities was sequentially investigated.Parameters such as organic modifier type and concentration, pH, tempera- ture and column type were studied. Experimental Reagents Tipredane metabolites and related substances were supplied by Fisons Pharmaceuticals Division (Loughborough) . Methanol (MeOH) (gradient HPLC grade) , acetonitrile (MeCN) (Far UV HPLC grade) and tetrahydrofuran (THF) (HPLC grade) used as the mobile phase modifiers and potassium dihydrogen phosphate (AnalaR grade) and orthophosphoric acid (85 .O%) * To whom correspondence should be addressed. (AnalaR grade) used to prepare pH buffers were supplied by Fisons Scientific Equipment Division (Loughborough). Water was purified by an Elgastat spectrum RO system supplied by Elga (High Wycombe). Chromatography HPLC analyses were performed on a HP1090M supplied by Hewlett-Packard (Stockport) and a Beckman Gold supplied by Beckman (High Wycombe).Both instruments possess diode array UV detection, column ovens and binary solvent delivery systems with solvent switching which allowed the use of two further solvents. General Chromatographic Conditions Unless otherwise stated chromatography was performed with a Waters Nova-Pak C18 (5 pm, 150 x 3.9 mm) column. Detec- tion was at 240 nm, based on the h,,, of tipredane. The flow rate was 1.5 ml min-' and the oven temperature held at 40 "C. Isocratic chromatography was performed unless otherwise stated. The suppliers of the columns used are shown in Table 1; all columns possessed stationary-phase packing of 5 pm except for the Hypersil column which possessed 3 pm material.The identity of the peaks in the chromatograms was established by comparison of retention times with those of individually chromatographed authentic samples. Results and Discussion Choice of Mobile Phase Initial selection of solvent strength In order to allow mapping of the resolution surface it was necessary to establish organic modifier-water mixtures for each organic modifier which gave suitable capacity factors of between 4 and 10 for all components being studied. As the elution order of the tipredane metabolites and related sub- stances could only be partially predicted by lipophilicityANALYTICAL PROCEEDINGS, JANUARY 1993. VOL 30 47 FPL 66365XX and FPL 66364XX OH OH e0 / - - - - 0 OH &yo / - - 0 FPL 67526XX 111 Table 1 Manufacturers and suppliers of columns studied Column and dimension/mm Manufacturer Supplier Hypersil Excel ODS Shandon Nova-Pak C l x Waters Ultrasphere C18 (150 Beckman Sphcrisorb ODs2 Phase Spherisorb ODS 1 Phase Resolve Clx Waters p-Bondapak C l x Waters Suplex pKb 100 Supelco Lichrosorb RP E. Merck (150 x 4.9) Scientific (150 x 3.9) x 4.6) (150 x 4.6) Separations (150 x 4.6) Separations (150 x 3.9) (300 x 3.9) (250 x 4.6) (125 x 4.0) select B Hichrom, Reading Fisons Scientific Equipment, Beckman, High Wycombe Loughborough Fisons Scientific Equipment, Fisons Scientific Equipment, Fisons Scientific Equipment, Fisons Scientific Equipment.Supelchem UK. Saffron Fisons Scientific Equipment. Loughborough Loughborough Loughborough Loughborough Walden Loughborough considerations and it has been demonstrated by Snyder et al.that suitable conditions for isocratic separation can be calcu- lated from the elution of the compounds under gradient conditions, initial chromatographic data for the compounds was established by chromatography using the tipredane related substances screen.2 The chromatogram obtained is shown in Fig. 1 and the identities and retention times of the peaks are given in Table 2. Only partial separation was achieved by this method with the compounds eluting as three pairs of poorly separated peaks. The compounds fall into three classes by polarity, the relatively polar sulfoxide diastereoisomers FPL 66365XX and FPL 66364XX, the moderately polar FPL 67526XX and FPL 66366XX and the relatively non-polar compounds I1 and 111.In general, chromatographic separations can be categorized into three classes as follows: (1) compounds which have very similar lipophilicities for which resolution is most difficult; (2) compounds with widely different lipophilicities which are easily resolved but will require gradient chromatography to obtain elution within a reasonable time scale; (3) compounds with moderately different lipophilicities which are readily separated by isocratic chromatography. The critical sepa- rations are those from categories (1) and (2) as these will determine the over-all nature of the method which is developed. From the results using the tipredane related substances screen it can be seen that the separation of FPL 66365XX from FPL 66364XX, FPL 67526XX from FPL 66366XX and of compounds I1 from 111 all fall into category (1).That of the first pair from the third pair falls into category (2), whereas the separation of the first pair from the second pair falls into the category (3). Approximate acetonitrile-water-buffer mixtures were iden- tified for each group of compounds in category (1) by reference to retention behaviour on the tipredane related substances 400 t FPL 66366XX I 0 1 I I I 5 10 15 20 25 Time/m i n Fig. 1 Chromatogram of tipredane metabolites and related sub- stances using tipredane related substances screen.’ Chromatographic conditions as described in the experimental section. Mobile phases A (0.025 mol dmp3 KH2POI) and B (0.025 mol dm-’ KH2POI in 65% v/v acetonitrile); a linear gradient was run over 20 min from 10 to 95% mobile phase B; then the eluent composition was held for a further 10 min. Each peak corresponds to approximately 1-2 pg loaded on to the column Table 2 Retention times (R.T.) of analytes on the tipredane related substances HPLC screen,’ calculated and actual acetonitrile contents of mobile phase used for isocratic elution MeCN R .T. /min Compound typical calculated(%) I FPL 66365XX 5.86 12.1 FPL 66364XX 6.01 12.6 FPL 67526XX 6.95 15.3 FPL 66366XX 7.03 15.5 I1 11.04 27.3 I11 11.17 27.6 MeCN used( % ) 12.5 12.5 12.5 12.5 25.0 25.048 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 gradient',2 (Table 2). The pH of the mobile phase was found to have no effect on the capacity factor, therefore water was used to adjust the elutropic strength instead of a suitable buffer. Solvent type The selectivity of reversed-phase systems is highly dependent on the mobile-phase modifier chosen.From the work of Snyder,3 methanol, acetonitrile and tetrahydrofuran were chosen as the most appropriate solvents for further studies due to their different selectivities. As will be shown later, the resolution of the tipredane metabolites and related substances requires the use of modifiers with different selectivities. Initially, the proportions of organic modifier necessary to obtain similar retention times with methanol and tetrahydro- furan mobile phases (isoelutropic mobile phases) were estab- lished by reference to literature values for solvent strength.4 Where these mobile phases did not elute the compounds in the required capacity factor window (4-10) the mobile phases were adjusted until suitable capacity factors were obtained (Table 3).Isocratic optimization f o r category ( I ) separations Category (1) separations were studied using the statistical design for selecting an optimum mixture of three solvents based on that suggested by Glajch et al.s using mobile phases which gave suitable capacity factors (see Table 3). The statistical design takes the form of a two-dimensional lattice search (Fig. 2). From these seven experiments it was possible to conclude which organic modifier gave the best resolution, as determined by the critical response factor (CRF)' for each of the category (1) separations, and whether ternary or quaternary mobile phases would give any improvement in resolution.From the CRF values shown in Table 4 (values of 0 and ~0 indicate baseline separation and co-elution, respectively) it was deduced that for compounds I1 and I11 methanol was required for resolution, whereas for the diastereoisomeric sulfoxides FPL 66365XX and FPL 66364XX acetonitrile gave the best resolution under the conditions employed. Binary solvent mixtures were chosen as ternary and quaternary systems offered no advantage. As the lipophilicities of FPL 67526XX and FPL 66366XX are similar to those of FPL 66365XX and FPL 66364XX, aceto- Table 3 Calculated and actual tetrahydrofuran and methanol concentrations for mobile phases with similar capacity factors to acetonitrile based mobile phase THF( "/o) MeOH(%) Compound MeCN(%) Calc. Used Calc. Used FPL 66365XX 12.5 8.9 7.5 15.4 40.0 FPL 66364XX 12.5 8.9 7.5 15.4 40.0 I1 25.0 17.8 15.0 30.8 40.0 III 25.0 17.8 15.0 30.8 40.0 Fig.2 Two-dimensional lattice search for three solvents A, B and C. The position of the numbers represent the trilinear coordinates of A, B and C for the seven mobile phases investigated Table 4 CRF values from two-dimensional lattice search FPL 66365XX FPL 66364XX and I11 and Compounds I1 Organic modifier Organic(%) CRF Organic(%) CRF Acetonitrile 12.5 0.02 25.0 0.9 Methanol 40.0 1.3 40.0 0.0 THF 7.5 Co 15.0 1.4 Acetonitrile-Methanol 6.3-20.0 0.08 12.5-20.0 0.0 Acetonitrile-THF 6.3-3.8 ~0 12.5-7.5 0.2 Methanol-THF 20.0-3.8 20.0-7.5 0.0 Acetonitrile-Methanol- THF 4.2-13.3-2.5 ~0 8.3-13.3-5.0 0.2 nitrile at the concentration used for the separation of FPL 66365XX from FPL 66364XX was the organic modifier of choice on practical grounds.This would avoid a rapid change in solvent strength or selectivity which would give rise to baseline fluctuations. Baseline separation of FPL 67526XX from FPL 66366XX was achieved with this mobile phase so no further optimization was performed. Tetrahydrofuran did not possess any advantages over the other organic modifiers and, given its incompatibility with the PTFE HPLC tubing and its greater toxicity, no further work was performed with this solvent. Optimization of solvent strength In order to gain more detailed information before moving from these isocratic methods to a gradient method further investi- gations into the chromatographic behaviour of the compounds were made by studying the relationship between organic modifier strength and capacity factor (k').The effect of methanol concentration on Ink' for com- pounds I1 and I11 is shown in Fig. 3. Unexpectedly the separation improved with increasing methanol concentration. This phenomenon was investigated further by studying the effect of acetonitrile concentration on the separation and the results are shown in Fig. 4. Here, elution crossover can be seen at about 22% acetonitrile: below this level compound I11 eluted before compound 11, whereas above 22% the elution order was reversed. Elution crossover was a direct result of differences in the gradient (S) of the line, in the graph of In k' versus organic modifier concentration, for each compound. A number of factors may lead to this difference in S .The difference in chemical nature of the analytes may be one factor; for example, the sulfone group of compound I1 may participate in dipole-dipole interactions with the mobile phases, whereas such interaction may be absent for the keto group of compound 111. In addition, the decrease in molecular size of compound 111 4.5 I I I I I I i I l l I 24 26 28 30 32 34 36 38 40 42 44 46 MeOH (%) Fig. 3 Effect of methanol concentration on In k' for compounds I1 and 111. Chromatographic conditions as described in experimental section. Compound I1 (A) and compound 111 ( A )ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 49 19.5 24.5 29.5 34.5 MeCN (%) Fig. 4 Effect of acetonitrile concentration on In k’ for compounds I1 and 111. Chromatographic conditions as described in experimental section.Key as in Fig. 3 compared with compound I1 may be another contributing factor: similar effects have previously been reported in studies of an homologous series of alkanes and carboxylic acids on p- Bondapak CI8 columns.6 In an analogous manner FPL 66366XX and FPL 67526XX exhibited differing gradients of In k’ versus acetonitrile concen- tration (Fig. 5 ) . In the light of our experience with the non-C6 hydroxy analogues I1 and 111, the effect of methanol concen- tration on In k’ was also studied and the results shown in Fig. 6. It can be postulated that if it had been possible to use higher percentages of acetonitrile without elution on the solvent front or lower concentrations of methanol without excessively long retention times elution crossover would have been seen in these cases as well.When acetonitrile was employed the elution order of the C6 3.5 3 2.5 & 2 1.5 1 0.5 0 C - 9.5 15.5 MeCN (Yo) 21.5 Fig. 5 Effect of acetonitrile concentration on In k’ for FPL 66365XX, FPL 66364XX, FPL 67526XX and FPL 66366XX. Chromatographic conditions as described in experimental section, with the exception that a Hypersil Excel ODS column was used. FPL 66365XX (0), FPL 66364XX (O), FPL 67526XX (0) and FPL 66366XX (H) 3.5 1 I 18 20 22 24 26 28 30 32 34 36 38 40 42 MeOH (Yo) Fig. 6 Effect of methanol concentration on In k’ for FPL 66365XX, FPL 66364XX. FPL 67526XX and FPL 66366XX. Chromatographic conditions as described in experimental section, with the exception that a Hypersil Excel ODS column was used.Key as in Fig. 5 hydroxy derivatives was observed to be FPL 66365XX followed by FPL 66364XX, FPL 67256XX and FPL 66366XX (see Fig. 5 ) , whereas with methanol there was a change in selectivity; the sulfone derivative FPL 66366XX now co-eluting with the sulfoxide FPL 66365XX (see Fig. 6). The diastereoiso- meric sulfoxides FPL 66365XX, FPL 66364XX and sulfone FPL 66366XX exhibited comparable gradients for the plots of In k’ versus methanol or acetonitrile concentration (see Figs. 5 and 6) and hence for these compounds no change in selectivity was observed over the organic modifier concentration range examined. However, the gradient for the keto derivative FPL 67526XX was consistently smaller, which was also the case for the corresponding non-C6 hydroxy keto derivative (compound 111).This suggested that molecular size may be a vital parameter in determining the magnitude of the gradient. PH There was no significant change in retention time or selectivity of either the sulfoxide diastereoisomers (which possessed an ionic character) or the compounds FPL 67526XX and FPL 66366XX (which lacked ionic character) over a mobile-phase pH range of 2.5-7.5 compared with that of unbuffered mobile phases. Column Temperature The capacity factors of FPL 66365XX and FPL 66364XX were measured at different temperatures by the use of a thermostati- cally controlled column oven. The expected linear reduction in Ink’ was found.7 There was no significant change in the selectivity within the temperature range examined (30-70 “C). Column Type The chromatographic behaviour of the sulfoxide diastereo- isomers (FPL 66365XX and FPL 66364XX) was investigated on columns of varying silica pre-treatment and degree of endcapping (Table 5 ) .Standard isocratic conditions of 12.5% acetonitrile in water were used for all the investigations, a flow rate of 1.5 ml min-’ was employed for columns of 150 mm length and the flow rate of column of other dimensions was adjusted to maintain comparable capacity factors. The retention times of FPL 66365XX, FPL 66364XX and FPL 66366XX relative to the retention time of FPL 67526XX on various columns are shown in Table 6. (Compounds FPL 67526XX and FPL 66366XX acted as markers and proved that the columns investigated had similar efficiencies.) Factors such as YO carbon load and trace metals content of the stationary phases were considered, but the only parameter which appeared to be closely related to the resolution of the sulfoxide diastereoisomers was the degree of ‘endcapping’ or deactiva- tion of the silica surface.Silica based columns, primarily designed for use with basic compounds (Tables 5 and 6, rows 7-9), which had been deactivated to yield a stationary phase with a very low numbers of active sites, gave no resolution of the sulfoxide diastereo- isomers. In addition, silica based columns which had not been deactivated (Tables5 and 6, row 6) and therefore possessed Table 5 Manufacturers information on columns used in study Silica pre- Column number Column Endcapped treatment 1 Hypersil Excel ODS Fully No 2 Nova-Pak C18 Fully No 3 Ultrasphere C18 Fully No 4 Spherisorb ODs2 Fully No 5 Spherisorb ODSl Partially No 6 Resolve CI8 No No 7 p-Bondapack CI8 Fully Yes 8 Suplex pKb 100 No Yes 9 Lichrosorb RP select B No Yes50 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 Table 6 Effect of column selection on relative retention time and selectivity Column number 1 2 3 4 5 6 7 8 9 Column Hypersil Excel ODS Nova-Pak CIS Ultrasphere C18 Spherisorb ODs2 Spherisorb ODS 1 Resolve C18 y-Bondapak C18 Suplex pKb 100 Lichrosorb RP select B Selectivity for FPL 6636SXW FPL 66364XX 1.13 1.10 1.11 1.09 1 .00 1.03 1 .OO 1 .OO 1 .00 Relative retention times FPL 6636SXX 0.63 0.69 0.64 0.69 0.97 0.91 0.85 0.76 0.87 FPL 66364XX 0.71 0.76 0.71 0.75 0.97 0.94 0.85 0.76 0.87 FPL 67526XX 1 .oo 1 .OO 1 .OO I .00 1 .OO 1 .OO 1 .OO I .OO 1 .OO FPL 66366XX 1.11 1.14 1.09 1.10 1.10 1.16 1.12 1.09 1.18 large numbers of active sites or were only partially endcapped (Tables 5 and 6, row 5 ) and contained a substantial number of active sites gave longer retention times and little or no resolution of the sulfoxide diastereoisomers.However, the sulfoxide diastereoisomers were resolved using silica based columns which were classified as ‘fully’ endcapped, yet without silica pre-treatment, by the manufacturers (Tables 5 and 6 , rows 1-4). The method of deactivation of these types of stationary phases gives rise to a small number of isolated active sites which are probably responsible for the selectivity observed. It is possible that the selectivity of the ‘fully’ endcapped columns arose from interaction of the sterically hindered active sites at the column surface with the sulfoxide moiety.Silanophilic interactions have previously been reported to explain the unusual chromatographic behaviour of antibiotics8 and of dibenz0-18-crown-6-ethers.~ However, this is the first reported example of selectivity differences for diastereo- isomers being attributed to these silanophilic interactions. Further investigations into these types of interactions are in progress. As a result of the above findings baseline resolution of the sulfoxide diastereoisomers could be achieved by the use of a 3 pm Hypersil Excel ODS column, which possessed the required characteristics, i.e. , ‘fully’ encapped and no silica pre- treatment, to provide the selectivity and give sharper peaks due to its higher efficiency compared with the other columns investigated.Selective Multisolvent Gradient Elution Method The final stage in the method development for the analysis was linking the conditions appropriate for each of the category ( 1 ) separations (compounds with very similar capacity factors) by a gradient to facilitate the category (2) separation (compounds with very different capacity factors) which was achieved by a simple linear gradient between the two sets of conditions. The resultant chromatogram is shown in Fig. 7. By definition the category ( 3 ) separation in acetonitrile based mobile phase (compounds with moderately different capacity factors) was readily achieved utilizing these conditions.Baseline separation of all six components was achieved within a satisfactory run time of 36 min. Conclusions Several interesting phenomena were observed during this investigation which re-emphasize the complex nature of the interactions which occur in reversed-phase HPLC. Firstly, the interaction of the sulfoxides with the stationary phase high- lighted the importance of column endcapping and silica pre- treatment in determining selectivity; secondly, the elution crossover which occurred showcd the importance of solvent strength in determining selectivity; thirdly, the literature values used to obtain isoelutropic mixtures did not give results in good agreement with experimental data, due to the importance of specific solute-solvent interactions in deter- 300 m z X $ 200 + :: 2 100 C m 0 Fig.7 10 20 30 Tirne/rn i n Chromatogram of tiorcdane metabolites and related sub- - stances on the selective multisolvent gradient elution method. Chro- matographic conditions as described in the experimental section. Hypersil Excel ODS 3 ym (150 x 4.9 mm) column was used. Mobile phases A (12.5% v/v acetonitrile in water) and B (95% v/v methanol in water), the mobile composition was kept at 0% B for 20 min then a linear gradient was run over 30min up to 100% B; then the composition was held for a further 5 min. Each peak corresponds to approximately 1-2 pg loaded onto the column mining retention times. Given the latter two observations, the relationship between In k’ and organic modifier concentration for each solvent and for each solute should be determined before selecting the mobile phases in the two-dimensional lattice optimization.This relationship can be more rapidly estimated for each organic modifier by comparison of capacity factors using two or more isoselective gradient elution methods with different rates of change of solvent strength. ‘O A selective multisolvent gradient elution method, where solvent strength, selectivity and composition were simulta- neously varied during the analysis, was developed using a systematic modular approach which relied to a greater extent upon analytical knowledge and interactive decision making than on fully automated optimization techniques. This resulted in baseline separation of the analytes of interest. The transfer from isocratic to gradient analysis was made easier by the greater over-all understanding of the separation obtained and because specific groups of compounds which were difficult to separate were identified.No optimization software was used and only a binary solvent delivery system was required. The method used requires the availability of individual groups of the compounds of similar capacity factor to allow isoelutropic investigation of areas of the solvent strength/ selectivity domain. In this particular application the individual compounds were also required for identification purposes, as the UV spectra of the individual compounds did not differ sufficiently for positive identification. The authors thank C. Thomson, who synthesized the com- pounds used in this investigation; V. Gerdelat and P.C. Downing for technical assistance; and R. J. Lewis for valuable discussions.ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 51 References 6 Snyder, L. R., Dolan, J. W . , and Gant, J. R., J . Chromatogr.. 1979, 165, 3. 7 Eucrby, M. R., Hare, J . , and Nichols, S. C., J . Pharm. Hiomed. Anal.. 1992, 10, 269. Snyder, L. R., J. Chromatogr. Sci.. 1978, 16, 223. 8 Intcrscience, New York, 1988, p. 136. togr., 1980, 199, 57. 1 2 3 4 Meyer, V. R., Practical HPLC Method Dcvetopment, Wiley- 9 5 Glajch, J . I,., Kirkland, J . J . , and Squire, K. M., J . Chroma- 10 Tanaka. N., andThornton. E. R., J. A m . Chem. Soc., 1977.99, 7300. Horvath, C., Melander, W . , and Molnar. I . , J. Chromatogr., 1976, 125, 129. Moats. W. A., J . Chromatogr., 1986. 366. 69. Nahum, A., and Horvath, C., J .Chromatogr.. 1981, 203, 53. Quarry, M. A . , Grob, R. L., and Snyder, L. R., Anal. Chem., 1986, 58, 907. Studies on the Use of Base-deactivated and Polymeric Stationary Phases in Reversed-phase High-performance Liquid Chromatography of Anthracyclines and Their Metabolites Glynis Nicholls, Brian J. Clark and J. E. Brown Pha rma ce u tica I Ch em is try, Sc h o o I o f Pharmacy, University o f Brad fo rd, Brad ford, West Yo rks h ire BD7 7DP The anthracycline group of antibiotics are amongst the most clinically important antitumour agents in current cancer chemotherapy; doxorubicin in particular has the broadest spectrum of activity of any known antineoplastic agent.’ The 4’-epimer epirubicin has recently been introduced into clinical medicine,* and has been reported to have equivalent antitu- mour activity to doxorubicin but with lower cardiac and systemic toxicity.We have recently described a method for the optimal separation of doxorubicin, epirubicin and their clinic- ally important metabolites using reversed-phase high-perfor- mance liquid chromatography (HPLC).3 In method develop- ment , formal simultaneous optimization strategies (the solvent selectivity triangle and factorial design approach) were used to define an area of factor space, within which the global optimal mobile phase conditions could be located using the modified simplex sequential optimization method. During this work, a complex mechanism of interaction between the stationary phase , mobile phase and anthracycline drugs and metabolites was observed. This included consider- able peak tailing when ionic forms of the drugs and metabolites were present and triethylamine was incorporated into the mobile phase as an endcapping agent to improve peak shape.During method development, complete resolution of all components was attained; however, only minor changes in mobile phase composition had significant effects on both resolution and retention times. Consequently, the robustness of the assay required further investigation, and in this work the function of the stationary phase in the separation was closely examined. This related in particular to the effect of the interaction of the anthracyclines with residual silanol groups present on the previously used silica-based stationary phase.Recent commercial developments in column technology have attempted to address the effect of residual silanols in reversed-phase HPLC separations, and alternatives to the traditional C18-bonded silicas have been introduced, especially for basic compounds. Of particular interest are the base- deactivated reversed-phase bonded silicas and polymeric column packings. In both types of stationary phase the effect of residual silanols on the chromatographic separation is neglig- ible, either because of the type of process used to manufacture the silica backbone, or through the use of non-silica materials. Therefore, these alternative stationary phases were used to investigate whether the over-all dependence of the assay method to changes in mobile phase composition could be improved by altering stationary-phase composition, whilst maintaining good resolution of the anthracyclines and their metabolites.Experimental The anthracycline drugs and metabolites assayed were doxoru- bicin, epirubicin, doxorubicinol, epirubicinol, 7-hydroxydox- orubicinol aglycone, 7-hydroxydoxorubicin aglycone, 7-deoxy- doxorubicin aglycone and daunorubicin (internal standard). The HPLC system consisted of: LKB 2150 dual reciprocating pump (LKB-Pharmacia, Uppsala, Sweden) , Hewlett-Packard 1046A fluorescence detector (Hewlett-Packard, Waldbronn, Germany) and Hewlett-Packard HP3394A computing integra- tor. For sensitive and selective detection fluorescence was used (detection limit for doxorubicin 1 ng ml-I), with excitation wavelength set at 480 nm and emission wavelength at 560 nm.Columns and Mobile-phase Systems Used (a) Five micrometre Hi-Chrom Spherisorb ODS 1, 250 X 4.6 mm i.d., with an optimized mobile phase: acetonitrile- 0.06 mol 1-’ citrate buffer (35 + 65 v/v), containing 0.05% (v/v) triethylamine, pH 4.6. ( b ) Five micrometre Hi-Chrom RPB Spherisorb, 250 x 4.6 mm i.d. Initial mobile phase was acetonitrile-0.06 mol I-’ citrate buffer (35 + 65 v/v), pH 4.6; this was modified by decreasing the acetonitrile content from 35% (v/v) to 30% (v/v) to improve chromatographic peak resolution. (c) Five micrometre Regis ‘Rexchrom’, 250 X 4.6 mm i.d. Initial mobile phase: acetonitrile-0.06 mol I-’ citrate buffer (35 + 65 v/v), pH 4.6. The optimized mobile phase as stated in ( a ) was also used. ( d ) Five micrometre Polymer Laboratories PLRP-S, 250 x 4.6mm i.d.Mobile phase: acetonitrile-0.06 mol I-’ citrate buffer (35 + 65 v/v), pH 4.6. Chromatographic peak efficiency was considerably improved in this experimental programme over the previous method3 by using 250 mm length columns instead of 150 mm. Peak efficiency values (theoretical plates) for all columns were calculated, relative to the parent drug doxorubicin. Resolution values were also calculated for all incompletely resolved peaks. Results and Discussion Conventional C18 reversed-phase packing materials often have residual silanol groups on the silica surface; these may interfere with the chromatographic separation of basic solutes and cause peak trailing and low efficiency. The glycosidic anthracyclines (Fig.1) contain an amine group which is protonated at all pH values at which the best conditions for resolution occur (pH (7). Interactions between the amine group and residual52 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 0 OH 'n Table 1 Stationary phases used and comparison of their effects on the separation of the anthracyclines and their metabolites. Mobile phase conditions as described in Figs. 2 and 3 Particle Nlplates Peak Resolu- Column size/nm m-'* shape tion RT Spherisorb ODS 1 5 26000 Good >1.5 RPB Spherisorb 5 31 000 Excellent 0.684 Regis 'Rexchrom' 5 6800 Poor 0.99 PLRP-S (polymer) 5 1000 Poor 0.77 drug, doxorubicin. separated pair of peaks in the chromatogram. modifier. * Peak efficiency ( N ) was calculated from the value for the parent T Resolution values were based on the resolution for the worst 4 Resolution was increased to R = 1.3 using 32% (v/v) organic Anthracyclinone ring system Daunosamine sugar residue (D) R1 R2 R3 R4 Doxoru bicin Epirubicin Doxoru bicinol Epirubicinol 7-OH Doxorubicinol aglycone 7-OH Doxorubicin ag lycone 7-Deoxy doxorubicin aglycone Daunorubicin D D D D OH OH H D -COCH20H H OH -CHOHCH2OH H OH -COCH20H OH H -CHOHCH20H OH H -CHOHCH20H - - -COCH20H - - -COCH20H - - -COCH3 H OH Fig.1 Structures of the anthracyclines 6 8 10 12 14 16 silanols were minimized in the original method (which used a partially endcapped Spherisorb ODS 1 stationary phase, with 7% carbon loading) by the addition of a short alkyl amine (triethylamine) to the mobile phase. An alternative approach is to use a stationary phase with a higher carbon loading (e.g., Spherisorb ODS 2, 12% carbon loading), although an endcap- ping mobile-phase additive is still required and the optimal component resolution conditions remain in a very narrow operating range.In addition, considerable variation in perfor- mance was found between different batches of Spherisorb ODS 1 stationary phases. Two alternative types of column packings were therefore investigated, which did not have the silanol effect problem and consequently may have a broader range of operating conditions. Base-deactivated Stationary Phases This type of stationary phase was developed primarily for the chromatographic assay of basic compounds, and is manufac- tured selectively to remove residual silanols. The packing material is not deactivated with a basic coating as the name implies, but is deactivated towards basic compounds.Hence, there is often no need for the inclusion of an endcapping agent in the mobile phase and good peak shape can be achieved, even with ionized basic compounds. Two different base-deactivated column packings were investigated, initially using the mobile phase conditions outlined in the previous paper3 but without the inclusion of triethylamine. Peak shape, efficiency and resolution were directly compared with those obtained using a 250 mm length Spherisorb ODS 1 column. 6 I I 1 1 I I I I 2 4 6 8 10 12 14 16 1 I 2 6 Hi-Chrom RPB Spherisorb By using the Hi-Chrom base-deactivated stationary phase, peak shape was improved for all of the anthracycline drugs and metabolites, and therefore peak efficiency was higher than that obtained using the Spherisorb ODS 1 stationary phase (Table 1).However, only partial resolution was obtained for two of the more polar metabolites (doxorubicinol and epirubi- cinol) compared with the baseline resolution attained pre- viously [Fig. 2(a), ( b ) ] . An improvement in resolution could, however, be achieved by decreasing the organic modifier content of the mobile phase from 35% (v/v) to 32% (v/v), but this resulted in an over-all retention time increase from 15.8 min to 18.2 min [Fig. 2(c)]. Baseline resolution of all peaks could only be achieved by reducing the acetonitrile content to 30% (v/v) of mobile phase (total run time 26 min). I I I I I 1 I I I 2 4 6 8 10 12 14 16 18 20 E I uti o n ti me/m i n Fig.2 HPLC separation of the anthracyclines and their metabolites using Spherisorb stationary phases: ( a ) Hi-Chrom Spherisorb ODS 1, with mobile phase 35% (v/v) acetonitrile-0.06 mol I-' disodium hydrogen phosphate-0.03 mol I-' citric acid containing 0.05% (vlv) triethylamine, pH 4.6; (b) Hi-Chrom RPB Spherisorb, with mobile phase as ( a ) ; ( c ) Hi-Chrom RPB Spherisorb, with mobile phase 32% (v/v) acetonitrile-0.06mol I-' disodium hydrogen phosphate- 0.03 mol I-' citric acid. pH 4.6. All columns were 250 x 4.6 mm i.d. stainless steel, packed with 5 pm size material. Flow rate was 1 ml min-', excitation wavelength was 480 nm, and emission wave- length was 560 nm. Key to peaks: 1, doxorubicinol; 2, epirubicinol; 3, 7-OH doxorubicinol aglycone; 4, doxorubicin; 5 , epirubicin; 6 , 7-OH doxorubicin aglycone; 7, daunorubicin; 8, 7-deoxy doxorubicin agl yconeANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 53 Although over-all resolution was not obtained without modification of mobile phase composition, the peak profile was considerably improved, which may significantly enhance sensitivity of detection in biological samples.However, it appears that the separation of the more polar anthracycline metabolites on Spherisorb ODS 1 packing material is partially dependent on secondary interactions with the residual silanol groups. Hence the ‘total’ endcapping found in this base- deactivated column resulted in poorer resolution of doxorubi- cinol and epirubicinol. This can be overcome by decreasing organic modifier content, but the over-all run time is increased, primarily because of the longer residence time in the column of 7-deoxy doxorubicin aglycone, the most lipophilic solute.Regis ‘Rexchrom ’ This base-deactivated stationary phase gave extended reten- tion times for all the anthracyclines when compared with the Spherisorb ODS 1 or the Hi-Chrom RPB Spherisorb packings. Peak shape was generally poorer and therefore efficiency was low (Table 1). However, with the exception of epirubicin and 7-hydroxy doxorubicin aglycone, all components were well separated [Fig. 3(a)]. In both of these base-deactivated stationary phases there is a possibility of some residual silanol effects. This was investi- gated by addition of 0.05% (v/v) triethylamine to the mobile phase, but in each case neither peak shape nor resolution were improved.Further optimization of the mobile phase was not attempted at this stage, but this may lead to some improve- ments in the separation. Thus the Regis column appears to be unsuitable for this particular application, as interaction between the anthracyc- lines and the stationary phase caused considerable peak broadening. Overall, the differences in resolution and efficiency observed between the Regis and Hi-Chrom station- ary phases probably reflect differences in manufacture or type of packing material used. Polymer based stationary phase (Polymer Laboratories Polymeric packings, recently developed for reversed-phase chromatography, have been promoted as a suitable alternative PL R P-S) to bonded-phase silica because of their stability over a wider pH range.They can be used ‘bare’ (e.g., polystyrene divinyl- benzene copolymers or PS-DVB) or derivatized with alkyl functional groups (e.g., Clx), and are reported to have different selectivity to silica-based columns. Additionally, polymer-coated or polymer-bonded silica materials are avail- able. The polymeric stationary phase used in this study was a PS-DVB copolymer. Using a mobile phase composition of acetonitrile- 0.06 mol I-’ citrate buffer (35 + 65 v/v) at pH4.6, the chromatographic peaks were generally poorly shaped, with peak tailing and incomplete resolution of four of the eight anthracycline components [Fig. 3(b)]. Nevertheless, retention times (R,) were comparable to the Spherisorb ODS 1 column, with the exception of the least polar metabolite 7- deoxy doxorubicin aglycone (R, >30 min), but plate efficiency was low (Table 1).The high retentive properties of the column, especially for the lipophilic metabolite, suggests that considerable hydro- phobic interaction occurred between the anthracyclinone ring system of the anthracyclines and the stationary phase. From this brief study, it can be concluded that ‘bare’ polymer-based stationary phases may be of less value than silica-based stationary phases for this application. Conclusions The base-deactivated stationary phases enable good resolution and peak shape to be attained for the anthracyclines under optimal conditions, providing there are only minimal hydro- phobic interactions of the anthracyclinone nucleus with the stationary phase. In this respect, the Hi-Chrom RPB column was found to be preferable to the Regis column, and it could be used as an alternative to the Spherisorb ODS 1 column if slightly longer run times were acceptable. The polymer column investigated was unsuitable for this application under the conditions chosen due to its high retentive properties for the anthracyclinone ring system. This was especially evident for the most lipophilic metabolite 7-deoxy doxorubicin aglycone. However, other less hydrophobic polymeric packings (e.g., alkyl-derivatized polymers) may be of use for this assay. With all the column packings investigated, the dependence of the 1 3 Inj. II II I I I I I 1 I I I I 1 I I Q, E O 2 4 6 8 10 12 14 16 18 20 22 24 26 28 .- 2 (b) 3 2 a, v) 6 3 U - 8 / - 7 I I I 1 I I I I I I I I I I I I I 1 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 Elution time/min Fig. 3 HPLC separation of the anthracyclines and their metabolites using: ( a ) base-deactivated stationary phase (Regis ‘Rexchrom’), with mobile phase 35% (v/v) acetonitrile-0.06 mol I-’ disodium hydrogen phosphate-0.03 mol 1-’ citric acid, pH 4.6; (b) polymer stationary phase (Polymer Laboratories PLRP-S), with mobile phase as (a). For other chromatographic parameters see caption to Fig. 2. Key to peaks as for Fig. 254 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 assay method to changes in mobile phase composition was not significantly improved. This study has demonstrated the influence of residual silanols present on silica-based reversed-phase HPLC supports on the retention of the anthracyclines, and their role in achieving baseline resolution of all components on a Spheri- sorb ODS 1 column. However, in considering alternative stationary phases, it is necessary not only to take into account potential drug-silanol interactions, but also the effect of interactions between the anthracyclinone ring system of these drugs and the hydrophobic components of the stationary phase. The authors thank the Science and Engineering Research Council for their support of this work, Farmitalia Carlo-Erba (Italy) for the donation of drugs and metabolites, and Hichrom Ltd. (Reading, UK) and Regis Chemical Company (Illinois, USA) for their donations of base-deactivated columns. References 1 2 3 Black. D. J . , and Livingston. R . B . , Drugs, 1990, 39, 652. Mross. K . , Maessen, P . , van der Vijgh, W. J . F., Gall, H . , Boven, E., and Pinedo, H . M., J . Clin. Oncol., 1988, 6 , 517. Nicholls, G., Brown, J . E.. Clark, B. J . . and Crawford. S. M., Anal. Proc.. 1993, in the press.