Preparative high-performance liquid chromatography purification of polyunsaturated phospholipids and characterization using ultraviolet derivative spectroscopy

188,136-141(1990)

ANALYTICALBIOCHEMISTRY

Preparative High-Performance Liquid Chromatography Purification of Polyunsaturated Phospholipids and Characterization Using Ultraviolet Derivative Spectroscopy’ Laura Montana

Received

L. Holte,’

Frederik

State University,

January

J. G. M. van Kuijk,3

Department

of Chemistry,

and Edward Bozeman,

59717

8,199O

A preparative reversed-phase HPLC system utilizing an isocratic mobile phase to purify up to lo-mg quantities of phospholipids is described. The method was developed to separate oxidation products of polyunsaturated phospholipids from intact, parent lipids. The method is useful for phosphatidylcholine and phosphatidylethanolamine on a preparative scale and for phosphatidylserine and phosphatidic acid on an analytical scale. Both intact phospholipids and oxidized phospholipids were monitored by absorbance at 206 nm. The oxidation products were simultaneously monitored at 234 nm where the intact phospholipids have only a very slight end absorption. Second-derivative uv spectroscopy proved to be extremely useful to identify the presence or to verify the absence of oxidation products in phospholipid samples. For autoxidized docosahexaenoic acid containing phospholipids, the absorbance maximum of diene oxidation products is 237 nm for the trans,trans (t,t) isomer and 246 nm for the cis,trans (c,t) isomers. Similarly, five classes of triene oxidation product stereoisomers have distinct absorbance maxima detected by second-derivative spectroscopy ranging from 269 to 292 nm. 0 1990 Academic PRESS. I~C.

There is great interest in the role of polyunsaturated fatty acids (PUFA)4 in biological systems (l-3). Highly i This work was supported in part by USA National Eye Institute Grant ROl-EY06913. ’ To whom correspondence should be addressed. 3 Supported by a grant from the Niels Stensen Foundation (The Netherlands). 4 Abbreviations used: PUFA, polyunsaturated fatty acids; 22:6w3, docosahexaenoic acid; (160) (22:6) PC, 1-palmitoyl-2-docosahexaenoyl-sn-glycero-3-phosphatidylcholme; (160) (22:6) PE, l-pahnitoyl-

136

Montana

A. Dratz

unsaturated fatty acids appear to play an important role in membrane structure and function (3-6). Dietary supplementation with highly unsaturated fatty acids has been reported to moderate certain degenerative diseases (7). A limiting factor in studying polyunsaturated phospholipids is their extreme sensitivity to oxidative degradation (8). The most readily oxidized lipids contain PUFA which have methylene-interrupted double bonds that facilitate free radical oxidation by abstraction of their bisallylic hydrogens. The oxidizability of PUFA is linearly dependent on the number of bisallylic methylenes present in a fatty acid (9). The most highly unsaturated fatty acid in biological systems contains six double bonds and is called docosahexaenoic acid or 22:603. Docosahexaenoic acid is typically esterified to the sn-2 position of a phospholipid glycerol backbone. The mechanism of free radical (R.) initiated autoxidation of 22:6w3 is shown in Fig. 1. Steps I and II show the abstraction of a hydrogen, forming a resonance stabilized free radical intermediate, and subsequent addition of oxygen. Regeneration of a free radical (R.) in step III allows propagation of the radical chain reaction. The autoxidative addition of one oxygen molecule to docosahexaenoic acid can yield up to 10 positionally isomeric hydroperoxides containing a conjugated diene (10). Oxidized fatty acids can lead to cellular damage (l&12) and may affect the properties of membranes (13,14). Oxidized fatty acids can further decompose to cytotoxic products (15,16). Synthetic phospholipids containing specific PUFA are often used to prepare model membranes for investi-

2-docosahexaenoyl-sn-glycero-3-phosphatidylethanolamine; (22:6) PS/PA, l-palmitoyl-2-docosahexaenoyl-sn-glycero-3-phosphatidylserine/phosphatidic acid; BHT, butylated hydroxytoluene.

(16:O)

0003-2697/90$3.00

Copyright 0 1990 by Academic Press, Inc. All rights of reproduction in any form reserved.

CHROMATOGRAPHIC

PURIFICATION

OF

I) eo@

---’

Ri

---

-

-

-

PHOSPHOLIPIDS

137

tion products have several characteristic features in the second-derivative uv spectra, and therefore, derivative spectroscopy was useful for rapid monitoring of crude or purified fractions to ascertain phospholipid purity.

rinivirulrji,a _ - - _ - . 1’ ‘\\

POLYUNSATURATED

+RH

MATERIALS

AND

METHODS

Reagents

r 02

II)

02 I

Zo” OO* III) floa

RH +R*

OOH

FIG. 1.

Mechanism of docosahexaenoate autoxidation: (I) An allylic proton is abstracted by a free radical initiator, R.. (II) An oxygen molecule can attack at one of two positions on the resonance stabilized radical intermediate. Addition of oxygen results in formation of a peroxy radical (-00. ). (III) The peroxy radical abstracts a hydrogen from another donor, regenerating a free radical (R.) which can propagate the chain process. The resulting hydroperoxide contains either a cis,truns conjugated diene, as shown, or a trans,truns diene.

gations of membrane properties. For most studies involving such membrane preparations, it is essential to work with lipid components which are intact and structurally defined. To avoid oxidation product-induced artifacts in biophysical studies or to investigate oxidative reactions as potential components of pathology, techniques are required to efficiently purify nonoxidized lipids and to separate and identify oxidation products. The HPLC method of van Kuijk et al. (17) developed for separating intact and oxidized phosphatidylcholine on an analytical scale was extended to purification of phospholipids with other head groups and modified for use on a preparative scale. Intact phospholipids absorb light only in the far uv, near 200 nm, so many common liquid chromatography solvents cannot be used with uvabsorption monitoring. The methanol and 0.1% ammonium acetate mobile phase used here is sufficiently transparent to permit detection by phospholipid absorbance at 206 nm. The isocratic system employed avoids column reequilibration between runs and allows for rapid separations. Derivative uv spectroscopy was used to characterize the presence or absence of oxidation products in samples before and after HPLC purification. Derivative techniques applied to uv spectroscopy greatly enhance resolution and contribute to an increase in detection sensitivity for minor oxidized components compared to standard, zero-order uv spectroscopy (18). Lipid oxida-

All solvents were HPLC grade from J. T. Baker Chemical Co. (Phillipsburg, NJ) and were used as received. HPLC-grade ammonium acetate was from Fisher Scientific (Fair Lawn, NJ). The mobile-phase solvent containing 0.1% ammonium acetate was filtered through a 0.5-pm Millipore Type FH (Millipore Corp., Bedford, MA) filter prior to use. 1-Palmitoyl-2-docosahexaenoylsn-glycero-3-phosphatidylcholine [ (16:0) (22:6) PC], 1 - palmitoyl - 2 - docosahexaenoyl - sn - glycero - 3 - phosphatidylethanolamine [ (16:O) (22:6) PE], and a l-palmitoy1 - 2 docosahexaenoyl - sn - glycero - 3 - phosphatidylserine/phosphatidic acid mixture [ (16:O) (22:6) PS/PA], were obtained from Avanti Polar Lipids (Birmingham, AL). Preparative High-Performance Liquid Chromatography Chromatography was performed on a system composed of two Altex 1lOA pumps and a Beckman 210A sample injector. Initial separations were performed with a 20-~1 sample loop on a 4.6 X 250-mm Alltech analytical column operating at a flow rate of 1.5 ml/min. Preparative separations were performed with a 2000-~1 sample loop on a 22.5 X 250-mm Alltech preparative column with a 22.5 X loo-mm guard column at ambient temperature. All columns were packed with Alltech Adsorbosphere HS Cl8 7-pm reversed-phase material. This is a high surface area packing with 60-A pore size and 20% carbon loading. A 100% methanol mobile phase containing 0.1% (w/w) ammonium acetate was delivered with two pumps, each at 8 ml/min for a combined flow rate of 16 ml/min. A typical preparative run took 30 min. The column effluent was passed through two detectors in series. The first was a Hitachi Model 100-10 variablewavelength uv absorbance detector operating at a wavelength of 234 nm. The second detector was an LKB 2238 Uvicord SII uv monitor equipped with a 206-nm interference filter. The outputs of both detectors were displayed on a Hewlett-Packard 7132A two-pen recorder. Second-Derivative uv Spectroscopy Ultraviolet absorbance and derivative spectroscopy were performed on a Shimadzu UV-3000 recording spectrophotometer. Fractions were collected from the HPLC, transferred to a quartz cuvette, and scanned from 190 to 300 nm usually versus a mobile-phase blank. Zero-order absorbance scans were typically recorded at

138

HOLTE,

r

3.

VAN

KUIJK,

1

.o

IA tt0.01 I hrt 11 B

.5.

i

t

.o.

.5.

t

i

/

A

o- L

-q

120

240nm260

280

AND

DRATZ

than the native, intact lipids. The oxidized lipids could therefore be separated from the intact lipids by reversedphase HPLC. Figure 4 shows a reversed-phase HPLC separation of (16:0) (22:6) PC from its oxidized products. The oxidation products were monitored at 234 nm in Fig. 4B and comprise a complex mixture of isomeric species that elute between 4 and 18 min. The intact, native phospholipids were monitored at 206 nm in Fig. 4A and elute at 23 min. A freshly purified (16:0) (22:6) PC fraction was collected from the HPLC and the methanol was evaporated to a small volume with argon. A small amount of butylated hydroxytoluene (BHT) was added (4 pg) to obtain a lipid to BHT ratio of 5OO:l. The sample was rechromatographed as shown in Figs. 4C and 4D. The single sharp peak at the solvent front is due to ammonium acetate which was concentrated from the eluant obtained during the initial purification. The small peak at 7 min is due to BHT added as an antioxidant. There is little absorbance in the 234-nm trace (Fig. 4D) due to the removal of the oxidized components by HPLC. The nonoxidized lipids monitored at 234 nm also produced a small peak at 23 min due to a slight end absorption from the strong 206-nm peak.

300

FIG. 2. (A) Zero-order uv spectra and (B) second-derivative uv spectra of partially autoxidized (16:O) (22:6) PC. A range of 2.5 AUFS was used for Trace A and a range of 0.1 AUFS for Trace B. Peaks of oxidation products are marked by arrows and correspond to secondderivative minima from left to right at X = 237,246,269,281,284,289, and 292 nm.

2.0-2.5 absorbance units full scale while a range of 0.1 absorbance unit was used for the second-derivative spectroscopy. This expanded scale for the derivative spectra provided a more sensitive enhancement of features of interest in the region 230-290 nm even though the derivative of the peak at 206 nm was off scale.

eo.olA

1

B

I RESULTS

Intact polyunsaturated phospholipids are colorless, but are observed to turn yellow if they oxidize and deteriorate. The yellow color results from conjugated oxidation products (see Fig. 1) whose uv-absorption peaks tail into the visible region and absorb blue light. Figure 2A shows a uv-absorption spectrum of rather heavily oxidized, yellowed (16:O) (22:6) PC with absorption maxima at 206,234, and 275 nm. Figure 3A shows an absorption spectrum of a colorless, freshly purified (16:O) (22:6) PC which retains the 206-nm peak but is essentially devoid of absorption in the region 230-290 nm. The oxidized products of the polyunsaturated phospholipids (such as those shown in Fig. 1) are more polar

FIG. 3. (A) Zero-order spectra of a HPLC-purified AUFS was used for Trace

uv spectra and (B) second-derivative uv sample of (160) (22:6) PC. A range of 2.0 A and a range of 0.1 AUFS for Trace B.

CHROMATOGRAPHIC

0

IO 1

,

PURIFICATION

20 I

OF

:30

,

1

1.2A

JII ).2A

206 nm C

I

15A

I,I

234 nm

D

I IO

I

I

I

20

POLYUNSATURATED

139

PHOSPHOLIPIDS

cessful purifications could only be done on an analytical scale (ca. 1 mg per run) using the preparative column. Table 1 shows a summary of retention times for phospholipids containing the head groups mentioned for preparative and analytical columns. Traces of oxidation products could be recognized by characteristic “fingerprint” spectral features in the second-derivative uv spectra. Figure 2C shows a second-derivative spectrum of heavily oxidized (16:O) (22:6) PC. The position of the centroid minimum in a second-derivative spectrum corresponds to the peak maximum in a conventional zero-order spectrum. Arrows in Fig. 2B mark absorption peaks of oxidation products of interest. The absorption peak centered at 234 nm has a shoulder at 242 nm, both of which emerge as minima in the second-derivative spectrum. The peak centered at 275 nm has spectral features and second-derivative minima at 269,281,284,289, and 292 nm. These spectra were taken in the absence of the antioxidant BHT but if BHT were present it would produce minima at 220, 275, and 282 nm in the second-derivative spectrum (data not shown). Figure 3B shows a second-derivative uv spectrum of freshly purified (16:0) (22:6) PC. The second-derivative spectrum shows small features which also occur in the instrument baseline and do not occur at wavelengths characteristic of oxidized, polyunsaturated phospholipids. DISCUSSION

The HPLC procedure described rapidly and completely separates polyunsaturated phospholipids from their oxidation products. An important advantage was gained by using HPLC columns with high carbon loading Cl8 packing. This packing more strongly retained the nonoxidized fatty acid chains and increased their re-

;

Retention time (min.) TABLE

FIG. 4.

Preparative HPLC separation of (16:O) (22:6) PC from its oxidation products. Traces A and B were obtained by injecting 0.2 ml of methanol containing 10 mg of partially oxidized phospholipid. The uv absorbance at 206 nm monitored in Trace A shows the oxidation products and the nonoxidized precursor phospholipid. Trace B monitored at 234 nm shows only the oxidation products. Traces C and D show chromatograms obtained by injecting ‘7 mg of freshly purified (16:O) (22:6) PC with monitoring at 206 and 234 nm, respectively. The first two peaks in C and D correspond to the concentrated ammonium acetate at the solvent front and the added BHT.

HPLC purification of (16:0) (22:6) PE is very similar to that shown above for PC (data not shown). Both PC and PE could be purified on a preparative scale with the column employed using typical sample loadings of 10 mg per run. Due to low solubility in the methanol mobile phase of phosphatidylserine and phosphatidic acid, suc-

1

Retention Times for 1-Palmitoyl-Z-docosahexaenoylsn-glycero-3-phospholipids’ Retention Analytical Lipid head group Choline Ethanolamine Serine Phosphatidic acid

column

time

(min) Preparative

Oxidized’

Nonoxidized

l-7 l-5 1-5

11.5 9.1 9.5

4-18 4-18 4-10

23.6 23.0 21.2

l-5

6.0

4-10

15.1

’ Conditions as described b The oxidation products range of retention times.

Oxidized!’

column

under Materials and Methods. include many isomers which

Nonoxidized

elute

over

a

140

HOLTE,

VAN

KUIJK,

tention times whereas the more polar oxidation products experienced rather little interaction with the hydrophobic stationary phase. The high carbon loading on the stationary phase extends the separation distance between the oxidized and the nonoxidized lipids. Extended separation is an important feature when employing preparative HPLC where resolution obtained on an analytical scale often deteriorates on scale-up. van Kuijk et al. (17) previously reported that the addition of 0.1% (w/w) ammonium acetate to the methanol mobile phase greatly sharpens phosphatidylcholine peaks relative to a pure methanol mobile phase. This sharpening effect was confirmed in preparative HPLC on phosphatidylcholine and phosphatidylethanolamine. The HPLC procedure described here was also found to be useful for analyticalscale separation of a mixture of polyunsaturated phosphatidylserine, phosphatidic acid, and their oxidation products using the preparative column (see Table 1). The addition of BHT is useful for inhibiting oxidation of the purified lipids after HPLC purification and a lipid to BHT molar ratio of 5OO:l is recommended. BHT absorbs in the uv region of interest and although low amounts can be blanked out, uv spectroscopy is usually best performed on lipids of interest prior to addition of BHT. The BHT was added as soon as possible after pure lipid fractions are collected from the HPLC. Purified phospholipids should be stored under argon at as low a temperature as possible. A freezer temperature of -120 or -75°C is preferred. Once polyunsaturated phospholipids are purified, it is desirable to have a routine procedure that can easily and accurately verify sample composition without the need for chemical derivatization (17,19) or other extensive manipulation of these delicate compounds. Derivative uv-absorption spectroscopy was found to be very useful for quickly indicating the presence or absence of oxidized polyunsaturated fatty acids directly on effluent fractions collected from the HPLC. Standard zero-order uv spectra have relatively low sensitivity for detection of small amounts of oxidized lipids. However, in derivative spectroscopy small spectral features are accentuated, especially for second and higher derivatives (20). While derivatives serve to reveal more spectroscopic information, higher derivatives also tend to amplify noise which may obscure the desired information. It was found that the second derivative was the best compromise between resolution enhancement and increased noise for the compounds of interest. The derivative resolution is a selectable parameter with lower resolution leading to noise averaging. The spectrophotometer used in this study calculates the derivative by fitting a polynomial to seven data points. We found that a derivative resolution of 3.5 nm (0.5 nm per point) reproduced all the features of the spectrum and the highest sensitivity for traces of oxidation products

AND

DRATZ

was obtained with a 6.3-nm resolution. Freshly purified lipid samples were found to be devoid of detectable oxidation products, judged by the absence of characteristic fingerprint peaks in the second-derivative uv spectra (see Fig. 3B). The spectrophotometer baseline did show small second-derivative peaks but these did not occur in the pattern characteristic of oxidized polyunsaturated fatty acids. Weenen reported uv absorbance values for autoxidized linoleic acid of 232.5 nm for the purified t, t isomer and 236 nm for purified c,t isomers (21,22). A recent paper by Corongiu et al. employed second-derivative uv spectroscopy to distinguish between mixtures of t, t and c,t stereoisomers of conjugated diene fatty acid hydroperoxides in rat liver microsomes by their respective second-derivative minima at 233 and 242 nm (23). In the present study the t,t- and c,t-diene stereoisomers of autoxidized (16:0) (22:6) PC were attributed to second-derivative minima at 237 and 246 nm, respectively (left two arrows in Fig. 4B). Presumably, the 4-nm wavelength shift we observe compared to the shift that Corongiu et al. (23) observed is due to additional unsaturation in the 22~6~3 fatty acid used in our work. Discrimination of oxidation product stereoisomers also appears to occur among the conjugated trienes, where the broad peak centered around 275 nm consists of up to five detectable components as revealed by minima in second-derivative spectra. Using the rule that a cis conformation leads to longer wavelength absorption than the tram conformation (24), the following assignments were proposed: the peak at 269 nm is attributed to the t,t,t-triene; t,c, t and t,t,c isomers are proposed to absorb at 281 and 284 nm; while the 289- and 292-nm peaks are thought to correspond to c, t,c- and c,c, t-triene isomers. The cis isomers tend to have lower extinction coefficients than the tram isomers, so the relative peak heights are not expected to directly reflect the relative amount of each product. Further work would be needed to firmly identify the second-derivative minima of these different stereoisomers. The HPLC purification and uv spectroscopy characterization procedures described are rapid and straightforward and utilize equipment common to many chemical and biological laboratories. Purification and frequent monitoring are important to avoid oxidation product-induced artifacts in studies involving polyunsaturated phospholipids. REFERENCES 1. Salem, N., Jr. (1989) in Current Topics in Nutrition and Disease (Spiller, G. A., and Scala, J., Eds.), Vol. 22, pp. 109-228, A. R. Liss, New York. 2. Kunau, W.-H., and Holman, Fatty Acids, The American

R. T. (Eds.) (1977) Polyunsaturated Oil Chemists Society, Champaign,

IL.

CHROMATOGRAPHIC

PURIFICATION

3. Dratz, E. A., and Deese, A. J. (1986) in Health saturated Fatty Acids in Seafoods (Simopolous, and Martin, R. E., Eds.), pp. 319-351, Academic 4. Wiedmann, T. S., Pates, R. D., Beach, J. M., Brown, M. F. (1988) Biochemistry 27,6469-6474. 5. Deese, (1981)

A. J., Dratz, Bio&mistry

E. A., Dahlquist, 20,6420-6427.

6. Watts,

A., Volotovski,

Effects of PolyunA. P., Kifer, R. R., Press, New York. Salmon,

F. W., and

I. D., and Marsh,

OF

A., and

Paddy,

D. (1979)

M.

R.

Biochemistry

18,5006-5013. 7. Kromhout, Engl. J. 8. Frankel,

D., Bosschieter, E. N. (1980)

9. Cosgrove, 299-304. 10. van

E. B., and Coulander,

C. L. (1985)

N.

Med. 312,1205-1209. Prog.

J. P., Church,

Rollins,

Lipid

Res.

M., and Murphy,

R. C. (1984)

W. A. (1987) J. Lipid

Lipids

22,

Res. 25,

507-

517. 11. Marx, 12. Bulkley,

J. L. (1987)

Science

G. B. (1983)

235,529-531.

Surgery

14. O’Brian, P. J. (1987) (Chan, H. W.-S., Ed.),

141

15. Esterbauer, H. (1982) in Free Radicals, Lipid Peroxidation and Cancer (McBrien, D. C. H., and Slater, T. H., Eds.), pp. 101-128, Academic Press, London. 16. Dratz, E. A., Farnsworth, C. C., Loew, E. C., Stephens, R. J., Thomas, D. W., and van Kuijk, F. J. G. M. (1989) in Vitamin E: Biochemistry and Health Implications (Diplock, A. T., Machlin, L. J., Packer, L., Pryor, W. A., Eds.), Vol. 570, pp. 46-60, New York Academy of Sciences, New York. 17. van Kuijk, F. J. G. M., Thomas, D. W., Stephens, R. J., and Dratz, E. A. (1985) J. Free Radicals Biol. Med. 1,215-225. 18. Butler, W. L., and Hopkins, D. W. (1970) Photochem. Photobiol. 19. van Kuijk, F. J. G. M., Thomas, D. W., Stephens, R. J., and Dratz, E. A. (1985) J. Free Radicals Biol. Med. 1,387-393. 20. Fell, A. F. (1983) Trends Anal. Chem. 2,63-66.

21. Weenen, 22.

94,407-411. G. J., and Dratz,

23.

in Autoxidation of Unsaturated Lipids pp. 233-280, Academic Press, London.

24.

13. van Kuijk, F. J. G. M., Sevanian, A., Handelman, E. A. (1987) Trends Biochem. Sci. 12,31-34.

PHOSPHOLIPIDS

12,439-450.

19, l-22.

D. F., and Pryor,

POLYUNSATURATED

H. (1982) Ph.D. thesis, Chap. 2, pp, 20-41, University of Nijmegen, The Netherlands. Porter, N. A., Wolf, R. A., Yarbro, E. M., and Weenen, H. (1979) Biochem. Biophys. Res. Commun. 89,1058-1064. Corongiu, F. P., Banni, S., and Dessi, M. A. (1989) Free Radicals Biol. Med. 7, 183-186. Williams, D. H., and Fleming, I. (1980) Spectroscopic Methods in Organic Chemistry, p. 1, McGraw-Hill, London.