Development of High Performance Liquid Chromatographic Separations for the Purification and Analysis of Phospholipids and Related Liponucleotides-A New Class of Anti-AIDS Drugs

The experimental anti-AIDS glycerophosphatidic acid:nucleoside (sn-1/sn-2 diacylglycerol:dideoxynucleotide) drugs, 3'-azido-3'-deoxythymidine monophosphate diglyceride ( AZT-MP-DG) [ 1 J and 2', 3 '-dideoxycytidine monophospha te diglyceride (ddC-MP-DG) [2] were isolated and purified by the reversed-phase mode of high performance . liquid chromatography ( RPLC). The chromatograph"ic separation was based upon the glycerophospholipid moiety of the drugs and the detection on the nucleoside component. The separations were optimized on method development columns packed with the stationary phase to be used in the micro-preparative column and monitored by a UV detector. Fractions we re collected and analyzed for purity by analytical-scale high performance liquid chromatography [3] and by thin-layer chromatography [4].

iii ACKNOWLEDGEMENTS I am deeply grateful to Professor Phyllis Brown for her support. Her guidance and drive has kept me going especially in "rough times". I would also like to thank her for all the opportunities she has given me. Words are not enough to express my many thanks. This dissertation has been written in manuscript form. The material is presented in two sections. Section one containins five papers and section two is comprised the appendix.
The first manuscript is a comprehensive review of the HPLC methodology, used for the preparative isolation and analysis of phospholipids and liponucleotides. This paper will be submitted for publication.

The second manuscript is The Preparative Isolation and
Purification of Lecithin. This paper has been published in Journal of Chromatography, 517 (1990) 219-228 The third manuscri"pt is on The development of an HPLC analysis of an anti-HIV glycerophospholipid, 3  Glycerophospholipids have been used to synthesize novel pharmace~ticals and to aid in drug delivery. They are used as membrane models, natural emulsifiers, wetting agents, dietetics and cosmetics. Analysis of glycerophospholipids help predict or diagnose certain disease states and determine metabolic pathways.
Since glycerophospholipids play important roles in many biological functions, it is necessary to have efficient separation methods both on the analytical and preparative scale.
Two chromatographic methods used are thin layer chromatography (TLC) and gas chromatography (GC). For TLC silica has been used extensively to isolate and quantify glycerophospholipid classes.
To fractionate individual molecular species, however, reversed phase or silica modified with silver ion is required. TLC does not possess the high resolving power for a complex mixture of glycerophospholipids and detection methods are difficult.
Glycerophospholipids, because of their polar head groups, are not readily amenable for separation by GC since they do not possess adequate volatility. For GC, the head group must be cleaved and replaced with a non-pola~ moiety, such as a methyl group. However, GC is one of the best methods to determine the fatty acid profiles. In this review glycerophospholipid separations by TLC or GC will not be discussed and the reader is referred to the following texts and reviews (1)(2)(3)(4)(5)(6) HPLC on the preparative scale provides the means to purify large quantities of glycerophospholipids in a short period of time while minimizing degradation. On the analytical scale, HPLC can be used for subsequent analysis of the separated glycerophospholipids.

Sample Preparation
The ideal sample introduction would be to inject the sample without any prior cleanup, but this in reality not possible.
Some type of sample pretreatment is necessary. The basic extraction for lipids are the methods described of Folch, Lees and Sloane (15) and Bligh and Dyer (16). Due to their insolubility in acetone, glycerophospholipids can be precipitated out of solution ( 17). The Fol ch extraction and acetone precipi ta ti on procedures are described in reference (18).  The URI Beacon, Wednesday, October 1, 1969  The U.R.I. sailing team sailed in two meets this weekend, taking a second and a fourth. On Saturday U.R.I. finished second to M.I.T. 13 • 19 in an invitational here on Salt Pond. 'Tom Dykstra, Skip Whyte, John Telfeyan and Henry Bassett each won at least one race. Other scores Saturday were Merrimack 30, Coast Guard 32, S.M.U. and Nichols 45, .and Stonehill 50. Mike Fenton and Chris Wells crewed.
On Sund•Y the team sailed at the Coast Guard Academy and finished fourth behind M.IT., Coast Guard, and Har· vard. There w_ ere three divisions .sailing at all times, one of Ravens, one of K boats. ·and one of dingheys. U.R.I. us'ed the same four skippers as the previous day. The racing was very close with M.I.T. and Coast Guard winding up tied with 94 points each after 5 ·races in each division. M.l, T. :w-0n by having more first place finishes. Other scores were Ilarvard 106, U.R.I. 109, Yale 125, Tufts 13Q, and Stevens 160.
Football Coach, Ram Grid Squad Face Trouble URI's football season cer· tainly has not opened on a very happy note. Rhody fans Who are not satisfied with the team's record for the past few years are chanting "Good Bye Zilly." Is it really the fault of the coach that the team is not as strong as we would like? Possibly it is, but it takes more than a coach to make or break a team. A team is just what the name implies -a team .
.. It takes many people to make a team successful. A good scouting staff is necessary in order to recruit the best talent available. Could it be that our scouting staff · is inadequate? Wit~out teamwork and a desire to win a team is not a team. Could it be that ~here is not enough teamwork between the members of our squad? There may be many reasons why the team is loosing. It may not be the fault of the coach, but the reason why they are loosing should be found out and eliminated. In any case, things could be pretty rough for coach Zilly if the team does not take a turn for the better. Under the coaching of Jack Zilly the team has a seven year record of 21

Wins Intramural Run
Rick Davids of Sigma Nu ·retained his title as URI Intraural Cross Country champion in the annual race which ' was held ill rainy weather last-: Friday: Rick's time was 9:46.5 which was well off the record of 9:33.5 which he set in 1967. Rick also won the champion-&hips in 1968, which adds up to a three year sweep of the event.
A field of 76 runners turned out for the race in spite of the drizzling rain and soaking wet course. The team title went to Sigma Nu with 57 points. Sigma Chi was second with 76 points ·and Sigma Phi Epsilon WIJS third with 121 points.

CHROMATOGRAPHY Preparative HPLC Class Separations
Preparative HPLC can be defined as any HPLC separation in which the effluent containing the solutes is collected and subsequently used as a reference standard, as a product, as a precursor or for other types of analysis. Therefore, preparative separations may be performed on virtually any column independent of the size. Columns range from the traditional analytical dimensions (15-30 cm x 0.20-0.46 cm i.d.) to preparative scale (15-200 cm x >15 cm i.d.). The column size, purity requirement Glass (37,38) replaced hexane with isooctane for the  (33,37,38,41), although there has been a report of partial species separation by silica (22) Figure   8. Chromatogram for the separation of soybean PC into its molecular species on a semipreparative Excello Ultra Pac ODS column (150 x 10 mm).

42
PA (peak b), PI(peak d), PE(peak e,f) were eluted in the initial solvent while PS(peak g), PC(peak h) and LPC(peak j) were highly compressed ( Figure 9). The break in the baseline was due to the abrupt change in water content.
separations on silica were examined using neutral and acidic mobile phases (58).   The gradient system used was previously described (27). Fairly broad peaks were observed and most of the solutes eluted as partially resolved pairs. The authors identified the peaks, however there was one discrepancy. The LPC was resolved near the sol vent front, while LPC has been reported to be highly retained eluting last after the SPH (27,28,60). This change in elution order may be caused by the type of silica used or the LPC was incorrectly identified. ~-Glycerolysophospholipids were synthesized enzymatically from the corresponding parent glycerophospholipid (60) and were separated with a mobile phase previously described (27  injection. An indirect method to determine the amount of 1 ns l· s conversion to lyso derivatives with exposure to plasma oge HCl fumes and analysis with HPLC. The separation of subclasses is becoming increasingly important. The plasmalogens are important components of central nervous system membranes, blood cells and many other vertebrate tissues (85)(86)(87). The subclasses have to be converted to nonpolar derivatives prior HPLC fractionation. The polar head group(s) are cleaved off with Phospholipase C and then acetylated. Normal phase HPLC was used to fractionate bovine brain PE into subclasses for subsequent HPLC species analysis (85). The PE was isolated from the other glycerophospholipids by TLC and converted to acetylated derivatives prior to injection. The mobile phase, which permitted UV detection, was composed of cyclopentane-Hex-methyl-t-butyl ether-acetic acid (73:24:3:0.03). Baseline separation of the three subclasses was achieved within 16 minutes ( Figure 13). Myher et al. (86,87) fractionated PE and PC derived from hu~an erythrocytes and human plasma into subclasses after derivatization to trimethylsilyl ethers. The mobile phase was a binary mixture of Hex-IPA (99.7:0.3). The fatty acid profile . of each collected subclass was determined by GC.

Analytical Species Separations
The separation of molecular species of glycerophospholipids has been a formidable and difficult task. Species separations Reversed-phase, C 18 , (88)(89)(90)(91) and silver ion impregnated, Ag+, silica (92,93) Table 3 in reference 90. PI and PS, peaks which were too small to be detected in the examples shown, were not numbered.
a. Jungalwala et al. (91) reported the species separation of SPH from bovine brain and sheep and pig erythrocytes with CH 3 0H-5 mM KH 2 P0 4 , pH 7.4, (9:1). Although 10 peaks were observed for the bovine brain, 11 for the sheep and 12 for the pig erythrocytes not all the molecular species were completely separated because some species coeluted. The fatty acids were determined by collecting each peak and analyzing each fraction by GC.
Ion-pairing agents have been added to the mobile phase to improve species separations of soy bean PI (95). Several tetraalkylammonium phosphates were used, which differed in length of the alkyl group . The longer chained ion-pair reagents caused an increase in the k' values of the PI species. Also an increase in concentration of the ion-pair reagent produced an increase in k'.

Argentation Chromatography
Chromatography performed on a silver column is called argenta ti on chromatography. A different type of interaction process takes place between the molecular species and the silver impregnated silica. The separation is based upon a complexation between the double bonds in the hydrocarbon chains and the silver, thus the separations are based upon degrees of unsaturation. There are few reported HPLC argentation separations for glycerophospholipids (92,93). A silver(I)-loaded resin was packed into an analytical column and used to separate phosphatidyl acid dimethyl esters derived from egg PC ( 92 ). Molecular species with one double bond were not well separated from those with two but were baseline resolved from the species with three double bonds. The eluent for the separation was diethyl ether. This type of silver stationary phase has been recently reported for the fractionation of egg and soy bean PC (93).   glycerophospholipids from a number of sources. The linear ranges and limit of detection (LOD) will be presented. There will be some reference to mobile phases and columns used, when quantitation is influenced.

HPLC-RI
One of the first on-line detectors used for HPLC is the refractive index detector (RI). The RI, a nondestructive bulk property detector, has been considered to be the "universal to r" because it is capable of responding to all solutes. detec Therefore solutes not need to posses UV absorbing chromophores.
The only criteria of response is the solutes must possess refractive indices that differ from the mobile phase, since it is the mobile phase which is constantly monitored. Good quantitation can be obtained, although sensitivity is moderate.
Though isocratic elution must be used, the choice of solvents comprising the mobile phase is virtually unlimited. Solvents in which are opaque in the UV can be easily used. The RI is a simple detector but has many disadvantages. It is sensitive to temperature, flow-rate and pressure fluctuations. In addition gradient elution is from a practical point impossible. Long equilibration times are usually necessary to achieve baseline stability, especially at sensitive settings.

Class Detection
On the ana l ytical HPLC s cale, the RI has been used to monitor class and molecular species separations. Response for the species separation was not limited to the degree of 75 · as in UV. Refractive index detection was utilized unsaturat1on n iter the isolation of hydrolysis products from PC (100) .. to mo The fatty acid chain from the sn-2 position was cleaved with phospholipase A 2 . The free fatty acids (FFA), LPC and residual PC were isocratically separated with an ACN-Me0H-H 2 0 (50: 45:6.5) and although the mobile phase was UV transparent, an RI detector was used. The isolated FFAs were converted to methyl esters and separated by GC for distribution analysis. Since relatively large amounts of the hydrolysis products were produced, the RI This mobile allowed the RI detector to be compared an UV detector. In addition the RI was contrasted to the conventional TLC-Phosphate (TLC-P) method. Coefficient of variation (C.V.) of peak area measurements for standard solutions was less than 2.0% for the range, 100 µg to 20 µg. The HPLC-RI method displayed a much smaller C.V. than the TLC-P method. Some inconsistencies in purity arose when the HPLC-UV was compared to the HPLC-RI method. The inconsistencies r esulted from the differences of UV response due to the fatty acid compositions, whereas RI response was not influenced by the fatty acids.
_fil>ecies Detection An isocratic molecular species separation for egg PC was described by Porter et al . ( 88). A reversed-phase and a fatty acid column was used with mobile phases consisting of MeOH-H 2 0-CHC13 (100: 10:10) and Me0H-H 2 0-CHC1 3 (70:19:10)  Human lung and carcinoma tissue PC was converted to diacylglycerol acetates, which were separated into molecular species and quantified with RI and UV detectors (105). The unsaturated species produced a higher UV response with respect to the saturated species, whereas good RI response was obtained for the saturated species. The UV and RI chromatograms are illustrated in Figure

HPLC-LLSD
The mass detector, based on light scattering principles, was described by Charlesworth (108) and can be considered as an "universal" detector for HPLC. The construction, principles of operation and evaluation have been thoroughly discussed (67,70,(108)(109)(110)(111)(112) In order to obtain an adequate signal response and stable baseline, the LLSD must optimized with respect to evaporation temperatures and gas flow-rate which are independently controlled (109,110,112). The temperature of the nebulizer is critical to providing a homogeneous analyte size distribution.
The temperature of the drift tube controls the evaporation efficiency of the mobile phase, while the gas flow regulates the spreading of the solute droplets while being carried to the laser.

Class Detection
Linear (68,74,113) and power (69,73,117) fi~s have been described for LLSD calibration curves. The linear curves have been further modified to include log/ log plots (67,70,109).
The %RSD were less than 1% for retention times and ranged from 0.97 to 12.1% for the area response. Minor components, such as PS and LPC, were detected from a lipid extract from 10 mg of rat liver.
Plots of log area versus log mass injected produce linear curves (67,70,109). The ternary gradient system developed by Christie (68) was modified to contain 1 to 5 mM ammonium chloride or NH 4 0Ac. The ammonium chloride caused a noisy baseline, which became more stable with the NH 4 0Ac. The ammonium salts were changed to 0.5 mM serine buffered to pH 7.5 with ethylamine which was applied to the analysis of rat kidney and rat brain (71) and cereals (72). The buffered mobile phase did not increase the background noise significantly and the back pressure was relatively constant.

Species Detection
The LLSD has been employed to monitor species separations of PC (97,114) and PE (114) but mobile phase modifiers, choline chloride and KH 2 P0 4 , had to be eliminated. An UV detector was in with a LLSD to compare responses for PC and PE species series (1!4)· Although the UV detector was more sensitive than the SD in that more components were detected, the LLSD provided LL ' a more representative analysis since peak size is a function mass but did not detect minor components. Some peaks with the highest UV intensity were recorded as minor components with the LLSD hence the authors suggested the use of both detectors.   a rent solvents, was developed so that an UV detector could transp be utilized. A chromatogram with FID and UV detection of lipid es is illustrated in Figure 18 (124). The FID and UV had class ar able sensitivities, but UV quantitation was more prone to comp errors due to the low response for saturated glycerophospholipids.

species Separations
Egg yolk PC ( 103)  solutes, may then be plotted (Figure 19 b-h). A partial species · can be observed from the peak splitting. Eighty separation micrograms of rat brain glycerophospholipids were separated (Figure 20) and quantified. Although relatively large masses were injected, the authors predicated that with specific ion monitoring (SIM) the quantity of analyte used may be decreased.
In addition difficulty arose when two or species had the same molecular weight.
Therefore if each class was further fractionated into molecular species the fatty acids with the same molecular weight could be distinguished .   carbonyl groups which strongly absorb at 5.7 µm. Therefore the IR may be operated at a specific wavelength or can be programed to acquire a entire spectrum. In an early report, an on-1 ine method was applied to quantify free acids and mono-, di-and triglycerides (140  The sensitivity could be improved by improving the design of the post column mixing chamber.

On-Line Phosphorus Analyzer
The determination of lipid phosphorus has been the classical method of analysis for glycerophospholipids.
Glycerophospholipids are composed of approximately 4% phosphorus        (Fig. 1). Besides being a naturally occurring emulsifier and surfactant, lecithin is also of interest as starting material for the synthesis of novel anti-viral and anti-tumor drugs [2][3][4][5][6]. Since purified lecithin was needed for the synthesis of these drugs, we developed a method for the isolation and purification of lecithin from chicken egg yolk, where a large amount of lecithin is present.
Patel and Sparrow [10]  were developed by Jungawala et al. (11] with a mobile phase of acetoni trile -rnethanol-water ( 65: 21: 14). In 1987 this mobile phase, to which 2-propanol and trifluoroacetic acid were added, was used for preparative work [ 20]. Gradient elution with a mobile phase of hexane-isopropanol-water has also been reported for analytical separations (16,17,21,22] and for preparative work (25]. However, isocratic elution is preferable if large quantities of lecithin are to be isolated routinely because of the ease of operation and cost savings. Ultraviolet detection is   ug/uL mixture of the standards was analyzed using the method development column. The PE and LPE, which had retention t i mes 152 4 7 and 7.8 minutes respectively were eluted prior to of · 'thin (retention time of 15.6 minutes); the SPH was eluted le Cl afterwards at 23.8 minutes, as was the lysolecithin at 28.8 minutes (Fig. 2).
with the method development column a loading study was performed to determine the injection volume of the working solution which was optimal to obtain adequate selectivity and lecithin purity. Samples of 10,25,50,75 and 100 uL were injected and fractions were collected across the lecithin peak.
The fractions were analyzed by analytical HPLC to determine the purity of the lecithin. Peaks were characterized by coinjections with the standards. The lecithin standard was eluted at 9.4 minutes (Fig. 3).In addition, two types of blanks were analyzed; the mobile phase before it entered the column (Before Column) and after the column (After Column) for each separation system. The fractions were also characterized by TLC using molybdenum blue spray to visuali z e selectively the phospholipids and sulfuric acid to detect all organic compounds.
The optimal injection volume was 50 uL (2mg) of the working solution (Fig. 4A). Nine fractions (2 minutes each) were collected, concentrated and analyzed. Of the 9 fractions, 3-7 contained only lecithin with the desired purity.
When the method development system was scaled-up for each of the pre para ti v e column s , the 1 inear velocity was k e pt constant. In scaling-up the volumetric flow rate scales as the square of the column radius, the sample load scales as the  (Table II) [4,5) as molecular targets or potential sites for drug intervention. At present the only FDA-approved drug for the treatment of AIDS / ARC and early asymptomatic HIV infection [6] is the antiretroviral analog of thymidine, 3'-Azido-2' ,3'-dideoxythymidine (Azidothymidine, AZT) which inhibits reverse transcription (polymerase). AZT and other clinical dideoxynucleos ides ( eg. ddI and ddC) exhibit dose-1 imi ting toxicity [7,8] and drug resistant HIV are known. Further, clinical dideoxynucleosides exhibit less than ideal pharmacokinetics, as evidenced by short plasma half-lives of typically 30-60 minutes [8].

177
HPLC method for the analysis of AZT-MP-DG. This method, phase which was optimized for reproducibility, precision, linearity and resolution was used to determine the purity of AZT-MP-DG.
The AZT-MP-DG was purified by two different preparative chromatographic methods (13]. With the rapid scanning UV, the sepa ration was al so monitored at 220 and 208 for the purposes of peak characterization and to detect impurities that do not have a chromophore that absorbs in the 250-280 nm region of the UV. Therefore the low UV region must be used to detect these impurities. However, the CHC1 3 used to solubilize the AZT-MP-DG produced a large solvent peak at low wavelengths as is shown in the 3-D plot in Figure 3. t.61 6.00 7.60 10.00 12." 49 1"4.99 17."48 19.98 BEJENllON TIME (Min ) 186 II I 1 , S used throughout this study. The retention was not I<H2P04 wa influenced by decreases in salt concentration, but at en trations greater than 5mM, the retention increased. To cone eliminate splitting, each standard was "conditioned" with 1 mL of the mobile phase and a minimum amount of CHC1 3 was used for solubilization of the AZT-MP-DG.

UV spectral Characterization
The absorbance ratios at two wavelengths provides a means of determining the purity of a compound. Usually, the ratio is compared to a literature value for confirmation, but reference values are not available for the AZT-MP-DG. The values of three sets of ratios ( 267 / 220, 267 / 208 and 220/ 208) are listed in Table I. The values were consistent for 5 samples of AZT-MP-DG. Since the relative standard deviation was lowest for the ratios 267 / 220 and 267 / 208, they are the ratios which should be used to characterize AZT-MP-DG.
In addition to absorbance ratios, the peak purity was determined with the rapid · scanning detector. Three UV spectra, which are normalized, were taken at three points along the peak.   .,.
. dicates that only a single component is present.

oG in purity Determination
In order to detect trace contaminants in pharmaceutical analyses injections of high concentrations [16,17] and the use of high sensitivity detectors [18] Figure 6. Linear response (267 nm) of powdered (y=2176.9x-433.6, n=5). Chromatographic given in Experimental. 201 detector as shown in the 3-D display (Figure 8). An important feature in Figure 8, is the absence of a large solvent peak in the low uv region, because the AZT-MP-DG in each fraction was in the preparative mobile phase which consisted of CH 3 0H/ H 2 0.

AZT-MP-DG conditions
Therefore, it was possible to monitor the fractions at the low wavelengths without the interference from the solvent peak. A representative chromatogram of the pooled fractions at 208 nm is shown in Figure 9. It clearly illustrates the insignificant absorption due to the solvent. However at 208 nm, the noise levels and baseline drift are more pronounced; therefore, interpretation may be more difficult. For instance, small baseline fluctuations may be recorded as a peak, depending upon the sensitivity of the detector and the integration parameters used.
Absorbance ratios (Table III)   are not available, our results were consistent when our methods were used on two different batches of AZT-MP-DG. The AZT-MP-DG behaved in a linear manner. The analytical method described can be used to analyze other synthenic liponucleotides [13]. This reversed-phase method has been scaled to a preparative separation for the purification of liponucleotides [13]. chromatography [3] and by thin-layer chromatography [4].
purity of the recovered ~rugs based on UV and light scattering detection, and on TLC was greater than 99%. The purified compounds were isolated for studies on structure confirmation, physical, biophysical and formulation properties, and anti-HIV efficacy in culture [1,2].

INTRODUCTION
In the treatment of acquired immunodeficiency syndrome (AIDS), AIDS related complex (ARC) and early HIV infection, chemotherapeutic agents are designed to attack one or more stages of the replicative cycle of the human immunodeficiency virus (HIV) [5][6][7][8]. An antiretroviral analog of thymidine, 3'azido-2' ,3'-dideoxythymidine (Azidothymidine, AZT) which inhibits reverse transcription (polymerase) is presently the only drug approved by the FDA for the treatment of AIDS / ARC and early asymptomatic HIV infection [ 9]. However, AZT and other dideoxynucleosides ( eg. ddC and ddI) exhibit doselimiting toxicity and have relatively short circulating lifetime [10,11]. In an effort to increase serum half -lives and decrease toxicity and consequently to increase efficacy, a new group of experimental liponucleotide anti-AIDS drugs, originating from earlier work on anti-cancer liponucleotides [12][13][14][15], has recently been synthesized [1,2]. These drugs  (16)(17)(18).
Solvents used to elute the liponucleotides usually contain large amounts of chloroform. Since this halogenated solvent is a carcinogen, it is preferable not to use it for the preparative isolation and purification of a drug. In addition chloroform precludes monitoring by ultraviolet (UV) detection at wavelengths less than 240 nm. Since phosphatidic acid and other possible contaminants of the synthetic mixture absorb at wavelengths below 210 nm, UV transparent solvents must be used.
To eliminate the use of chloroform, which is often used with silica columns to purify compounds, a reversed-phase HPLC method was developed utilizing a UV transparent solvent system. The  16:0/18:lw9 (sni;sn-2) phosphatidic acid (PA) [1,2] [22,23]. In addition to KH 2 P0 4 , ammonium acetate (NH 4 0Ac) can be used as the salt in the . mobile phase [24]. The NH 4 0Ac at 1 mM gave the same k' values as those obtained with 1 mM KH 2 P0 4 • For purity analysis, aliquots of the collected fractiorts were injected onto the analytical columns. Since commercial standards were not a vailable for either compound, the liponucleotide peaks were cha racterized with UV detection using absorbance ratios and peak purity measurements. In addition to UV and TLC analysis, a light scattering .detector was used [25]. The rapid scanning UV detector was used to obtain absorbance ratios and peak purity values. Three wavelengths at 267, 220 and 208 nm were used to determine the ratios. The ratios at 267 / 220, 267 / 208 and 220/208 w.ere compared to ratios obtained on previously purified AZT-MP-DG. For multiple determinations the absorbance ratios were consistent, as shown in Table I   TUDE ~·NJ 23 9 MP-DG at 20S nm did not reveal any additional impurities.
Representative chromatograms of the pooled fractions at 2SO nm and 20s nm are shown in figure Sa and Sb. Based upon UV absorption the purity of the pooled ddC-MP-DG was 99.S%. The pooled fractions were also monitored with the rapid scanning UV detector. Absorbance ratios, listed in Table II,    N study naturally occurring phospholipids in egg yolk were used as model compounds and were separated on silica supports.
To monitor class separations on an analytical scale, re f r a c t i v e i n de x ( RI ) ( 4 , 5 ) and u 1 t r av i o 1 e t ( UV ) ( (11)(12)(13) or fluorescent ( 14)  Other detectors have also been used in the HPLC analysis of phospholipids. The flame ioni z ation detector (FID) has been described for quantitation of soy bean and synthetic phospholipids (15,16) and multispecies quantitation of phospholipids (17,18); howe v er it has not g a ined widespread use with HPLC. An electrochemical method based on tensammetry has recently been described for the determination of phospholipids in serum (8) An evaporative light scattering detector (ELSD), which is an universal transport detector, has been reported for HPLC (27)(28)(29). This detector is based on light scattering principles and has many adv antages. One of the major advantages is that it can be used to monitor high molecular weight, non-volatile compounds such as phospholipids (30)(31)(32)(33)(34)(35)(36)(37)(38), triglycerides (39) and carbohydrates (40). Quantitation is based directly on the mass of a substance injected. With the ELSD, solutes need not possess chromophoric groups but they must be non-volatile, a characteristic inherent in large molecules; however mobile phases must be v olatile. In a ddition the eluent can be UV opaque and gradient elution is easily used. The limitations of the ELSD are that the solutes are not recoverable and the 2 59 mobile phase can not contain buffers or salts which will precipitate in the nebulizer syringe, although trace amounts of these ionic substances have been used (34).
Infrared (IR) spectroscopy has also been used to monitor phospholipids after HPLC separation (41). The IR offers functional group characterization and quantitation.
Quantitation can be based on monitoring the absorbances at the wavelengths where the carbonyl and amide group ( s) absorb.
Since expensive deuterated solvents must be used in the mobile phase for direct coupling of the HPLC column to the detector, an off-line IR interface h~s been developed for HPLC, whereby deuterated solvents are not required (42,43).
Therefore the performance of the ELSD was evaluated and   (30,31,33,35,38). A mobile phase consisting predominantly of CHC1 3 was used with the ELSD because CHC1 3 is more volatile than mobile phases ordinarily used with UV detection. In addition phospholipids are more soluble in CHC1 3 rich solutions. However, the mobile phases commonly used with a UV detector can be utilized . if the gas flow-rate and exhaust and heater temperatures are optimized for these eluents ( 47). It has been reported that increased resolution can be obtained by the addition of strong acids (48,49), but when sulfuric acid or TFA (5-lOmM) was added to the mobile phase used with the ELSD, broad peaks and baseline instability resulted.
For previous work ammonium acetate in the mobile phase sharpened phospholipid peaks (44); however ammonium acetate could not be used with the ELSD because of precipitation of the acetate in the nebulizer tubing and syringe. Thus ammonium hydroxide was substituted (30). With 4mM ammonium hydroxide present in the mobile phase, each phospholipid eluted as a single sharp symmetrical peak except for the LPC which eluted as two peaks, probably due to a partial species separation.
The mobile phase for the AZT-MP-DG was previously developed for a reversed-phase system with UV detection (45). For the ELSD ammonium hydroxide was substituted for the potassium dihydrogen phosphate.
For the IR Transform a volatile eluent is also necessary.
Therefore a mobile phase of chloroform-methanol-water was employed. Trifluoroacetic acid (TFA) which is very volatile, was added to decrease band broadening (48,49), however, the peaks were broadened instead of sharpened.
For UV detection, ternary low UV transparent solvents have been used as the mobile phase. Two types of mobile phases have been described; acetonitrile-methanol-water (10) and hexaneisopropanol-water ( 14). In this st_ udy, the mobile phase of Jungalwala et al. (10) was modified for the analysis of the PC in fractions from preparative HPLC separations (44).
The mobile phase of chloroform-methanol-water ( 60: 40: 4) which had been used with the ELSD detector was used with the RI detector.

Evaporative Light Scattering Detection
For the ELSD the parameters which had to be optimized were mobile phase and nitrogen gas flow-rate, exhaust and heater temperatures. A fairly fast mobile phase flow-rate was used because of the high volatility of the chloroform. Consequently a high carrier gas flow-rate was required to insure adequate vapori z ation and minimize band broadening. The temperature settings of the heater and exhaust were critical. Low temperatures resulted in inadequate vaporization of the solvent causing baseline fluctuations and high temperatures vaporized the solvent too quickly. The temperature also affected the detector response by causing variations in solute droplet size (29). The r efore, du r ing the optim i zing process, the temperature and nitrogen gas flow-rate were adjusted with respect to the mobile phase flow-rate, in order to obtain the greatest response. Since the response of the ELSD does not depend upon chromophoric groups nor on the refractive index of the solutes, it is suitable for monitoring naturally occurring or synthetic phospholipids . The ELSD was operated at the most sensitive setting with negligible baseline noise even at negative attenuations on the integrator. Retention times and areas were very reproducible (Table I and Table II). The PE under these condit i ons eluted near the v oid v olume; thus it could not be detect e d with the UV detecto r due to the large solvent peak but could be reproducibly detected with the ELSD. The response of the ELSD was non-linear and a logarithmic function best fit the  (33). A linear log/log plot (R 2 > 0.99) was obtained with a non-zero intercept ( Figure 2). A linear equation was derived for each phospholipid (Table III). The working range for this study was 0.0969 (1.25 µg) to 1.6990 (50 µg) and a detection limit of 0.25 ug (PE), 0.50 ug (LPE, PC and SPH) and 1.0 ug (LPC) was obtained. Lower detection limits with the ELSD have been reported with gradient elution (30,33).
A standard mixture of the five phospholipids, consisting of 10 µG of each compound, was injected. The phospholipids were well resolved with excellent peak shapes ( Figure 3). The separation time was less than 20 minutes. The mass of each phospholipid in the mixture was calculated with the regression equations (Table IV). There was good agreement between the injected and the computed amount. The LPE had the largest deviation in the correlation between mass injected and mass calculated. It is postulated that the deviation was due to incomplete dissolution, since the LPE had the lowest solubility.
When a 1.18 µg / µL solution of the crude egg yolk phosphol ipids was injected, al 1 the major phosphol ipids we re well resolved ( Figure 4). The major phospholipids were PE and PC, while LPE, SPH and LPC were the minor ones. The percent of each phospholipid was calculated with the regression equations and there was good agreement with the literature values as shown in Table v (50,53). Since the phospholipid composition varies in different sources of eggs, the literature values must be considered to be general in nature.       After a preparative procedure for the isolation and purification of PC (44), the collected fractions were analyzed for purity. A representative HPLC-ELSD chromatogram shown in Figure 5 illustrates that only PC is present. The high peak purity was also confirmed with UV detection.
The ELSD was also useful in confirming the purity of AZT-MP-DG ( Figure 6). The ELSD conditions, nitrogen gas flow-rate and temperatures were adjusted to vaporize adequately the mobile phase which has been developed for the HPLC separation with UV detection, to maintain a stable baseline and to exhibit good response. The high methanol content (95%) was easily evaporated, although sensiti v ity was decreased due baseline drift. The compound had not been removed from the micro-preparative eluent which was methanol-lmM KH 2 P0 4 (95:5). Although a trace amount of KH 2 P0 4 was present in the aliquot injected, the response and pressure remained constant during the analysis indicating that the salt did not precipitate. No other solutes were detected demonstrating the high purity of the compound. The high purity was also found with the UV analysis (45,54). However, with UV detection, impurities which do not have a chromophore can not be detected.

Inf rared Detection
Infrared spectroscopy provides functional group characterization as well as quantitation. Direct coupling to an HPLC has been difficult, because of engineering problems and the limitations imposed by solvent absorption. An alternative off- line system has been developed in which the effluent does not enter the IR sample compartment, thus eliminating many of the problems associated with direct coupling (42,43). A 1:25 splitter was used to divide the effluent. At a flow-rate of 1.8 mL/min, 0.072 mL/min was delivered to the IR Transform and nebulized. The mobile phase was evaporated and the solutes deposited through a syringe tip onto a rotating Germanium reflective disk. The rota ti on angle of the disk at any given time was known and the speed of rotation controlled. Once all the solutes had been deposited onto the disk, it was placed in a reflectance accessory in the FT-IR sample chamber. The reflectance accessory also rotated the disk to any angle where a solute has been deposited and an IR spectrum was taken.
An injection of a mixture containing 40 µG each of PE, PC, SPH and LPC resulted in 1.6 µg of each analyte being deposited on the disk. The IR spectra (Figure 7 A-D) clearly show functional group absorptions; for example the carbonyl, C=O, stretch at 1750 cm-1 to 1650 cm-1 • Othe~ absorptions are due to the C-H stretch in the fatty acid chains at 3000 cm-1 to 2800 cm 1 , the phosphate group, P=O, at 1300 cm-1 to 1250 cm-1 and the covalent phosphate, P-0-C, at 1050 cm-1 to 970 cm-1 (51). The wave numbers for specific functional groups are in Table VI. For quantitative analysis the FT-IR can also monitor a particular functional group wavelength, which is analogous to a UV detector set at a specific wavelength. During a separation the ELSD was coupled in parallel to the IR Transform by connecting the waste  With the angle known, the disk when placed in the accessory in the FT-IR sample chamber was rotated to the exact angle where deposition occurred. The coupling with the ELSD eliminated the need to rotate the disk every few degrees and take a spectrum.

Ultraviolet Detection
Although ultraviolet detectors are very sensitive, the sensi ti vi ty for phospholipids is primarily dependent on the number of double bonds in the diglyceride moiety. Moreover, the distribution of the fatty acid moieties differ within each phospholipid class. Consequently, direct quantitation by UV of phospholipids with an unknown number of double bonds is difficult. At lower wavelengths which are nearer the wavelength maximum, high sensitivity is obtained but at the expense of high background due to the mobile phase~ A reported limit of detection (8) was 0.2 µg . for natural occurring phospholipids with a linear response ranging from 0.2 to 15 µG, whereas the limit was 2.5 µg for synthetic phospholipids. The lower detection limit for the naturally occurring phospholipids is due to the presence of a larger number of unsaturated fatty acids.
The lyso-derived phospholipids, LPE and LPC, are inherently more difficult to detect by UV because they lack a fatty acid chain ( Figure 1 c and D). Since the detector "sees" half the compound with respect to PE and PC, the UV response of the lyso derivatives is lower than for PE and PC, (52).
A solvent peak, due to the chloroform-methanol solution used to solubilize the phospholipids, was very pronounced at low UV wavelengths. The tail of the peak obscured the PE and LPE peaks which eluted near the solvent front. A solvent peak is usually absent with the ELSD since the mobile phase is volatilized.
Detection by UV has been primarily used to determine the purity of PC (44) and AZT-MP-DG (45) isolated from preparative HPLC separations. Figure 8 is a 3-D chromatogram of the pooled fractions containing AZT-MP-DG in the highest purity (>99%) from a preparative HPLC separation.

Refractive Index Detection
Phosphatidylcholine from egg yolk and soy bean has been monitored with a RI detector. The RI detector was compared with a UV detector and TLC for purity determination of PC (5). A major advantage of the RI detector is that the response is not dependent upon the degree of unsaturation as in UV. However the response is affected by lack of sensitivity and temperature and baseline instability. Refractive Index calibration curves of PC have been reported to be linear in the range of 20 µg to 100 µg (5). Initially, the RI detector was used to monitor PC analyses, but due to its lack of sensitivity small quantities of other phospholipids were not detected. The amount of PC relative to  296 the total phospholipids content can vary from 65-85% in egg yolk (50,53), while less than a few percent of other phospholipids may be present. A HPLC-RI chromatogram of crude egg yolk phospholipids, illustrates the lack of sensitivity of the RI detector for the minor phospholipids ( Figure 9). Although the PC was well detected, the PE and LPC were obscured by the solvent peaks and the SPH and LPC could not be detected.   17 (1984) 1857-1862 Elfakir, C., Lafosse, M. and Dreux, M., Optimization of High Performance Liquid Chromatographic Analysis with UV Detection: Light Scattering Detection to Establish the Coelution of UV and Non-UV absorbing Constituents. Journal of Chromatography, 513 ( 1990) 354-359 Ellingston, J.S. and Zimmerman, R.L., Rapid separation of gram quantities of phospholipids from biological membranes by preparative high performance liquid chromatography. Journal of Lipid Research, 28 (1987)