MAGNETIC RESONANCE IN MEDICINE 25, 120- I27 ( 1992)

Selective Suppression of Lipid Resonances by Lipid-Soluble Nitroxides in NMR Spectroscopy

KAI CHEN, * .j- NORBERT W. LUTZ, * JANNA P. WEHRLE, * JERRY D. GLICKSON* AND HAROLD M. SWARTZ *'?,$

* Division of NMR Research, Department of Radiology and Rudiological Sciences, The Johns Hopltins University School of Medicine, Baltimore, Muryland 21205 and tC'ollege of Medicine,

University of Illinois, Urbanu, Illinois 61801

Received January 2, 1991; revised May 22, 1991

The ability of lipid-soluble nitroxides to suppress selectively the peaks of lipid resonances in 31P, 'H, and "C NMR spectra was investigated in serum as part of studies aimed at using these contrast agents for magnetic resonance imaging and magnetic resonance spec- troscopy in vivo. Nitroxides are especially interesting potential contrast agents because they can reversibly be converted in cells to diamagnetic hydroxylamines, with conversion rates that are dependent on the redox potential and the intracellular concentration of oxygen; the characterization of nitroxide-dependent changes in NMR spectra may therefore be a useful means to measure oxygen-dependent redox metabolism in vivo. The fatty acid analogs, doxy1 stearates, suppressed the methyl resonance of choline and the methyl and methylene peaks of lipids in the 'H NMR spectra of serum samples. As a consequence, lactate peaks, which were not readily detected became clearly resolved and could be evaluated quantitatively. The "P resonance of phosphatidylcholine in the "P NMR spectrum was suppressed by 5-doxy1 stearate and 4-( N,N-dimethyl-N-hexadecyl)ammonium-2,2,6,6- tetramethylpiperidine-I-oxy1,iodide (Cat,,). In the I3C NMR spectrum, the resonances of the methyl groups of choline and the lipids also were broadened significantly by addition of 5-doxy1 stearate. Differential suppression of lipid resonances can be employed to facilitate quantitation of lactate. o 1992 Academic Press, Inc.

INTRODUCTION

This report is part of an ongoing effort to develop techniques ( 1 ) to affect selectively lipid resonances by lipid-soluble paramagnetic nitroxides in order to obtain contrast specifically in fatty tissue, (2) to enhance observation of underlying spectral features in NMR spectroscopy, and ( 3 ) to develop hypoxia-sensitive contrast agents for in vivo NMR. We have made our initial studies in serum because it is a relatively simple model system, and, as indicated below, there is a particularly interesting application in serum for this approach: the resolution of the proton spectrum of lactate.

The use of lipid-soluble paramagnetic contrast agents has the potential of selectively yielding high concentrations of the agent in particular sites in cells and tissue and should provide observable effects with modest amounts of contrast agents.

To whom correspondence and reprint requests should be addressed at Department of Radiology, Dart- mouth Medical School, 308 Strasenburgh Hall, Hanover, NH 03755-3863.

0740-3194192 $3.00 120 Copynght 0 1992 by Academic Press, Inc All nghts of reproduction in any form reserved

SELECTIVE SUPPRESSION OF LIPID PEAKS 121

The use of nitroxides is based on the wide range of chemical structures that are available for this class of contrast agents and their potential as metabolically responsive contrast agents ( 1 ). Lipid-soluble nitroxides can be converted reversibly to diamagnetic hydroxylamines in cells (2, 3 ) . The rates of reduction of lipid-soluble nitroxides are strongly dependent on the intracellular concentration of oxygen; profoundly hypoxic cells reduce nitroxides more rapidly than cells supplied with oxygen ( 4 , 5 ) . The rates of oxidation of hydroxylamines increase with increasing intracellular oxygen concen- tration up to 150 pM ( 5 ) . The characterization of nitroxide-dependent changes in NMR spectra could be used to measure the concentration of oxygen because the concentration of lipid-soluble nitroxides is affected by the concentration of oxygen in cells, if there is a rapid exchange of nitroxides between cells and lipids.

'H NMR resonances of low molecular weight metabolites in serum such as lactate, acetone, and acetoacetate often are not well resolved from broad overlapping resonances originating from protons of proteins and lipids. Spin-echo Fourier-transform techniques (6-8) can be used to suppress these broad overlapping resonances; however, the absolute peak intensities of the low molecular weight metabolites might be affected as well, especially when some relatively sharp resonances of lipid protons are suppressed. In this report, we show that lipid-soluble nitroxides can suppress lipid resonances selec- tively without decreasing signal intensities of the low molecular weight metabolites such as lactate. These effects can be used to make quantitative measurements of lactate.

MATERIALS AND METHODS

Horse serum was purchased from GIBCO Laboratories (Grand Island, NY). Li- poproteins were purified by ultracentrifugation in NaCl and KBr solutions ( 9 ) . Nic- otinamide adenine dinucleotide (NAD) , glycine buffer, lactate dehydrogenase, and lactic acid standard solution were purchased from Sigma Chemical Co. (St. Louis, MO). 5- and 12-doxy1 stearates, and 4-( N,N,-dimethyl-N-hexadecy1)ammonium- 2,2,6,6-tetramethyl-piperidine- 1 -oxyl,iodide ( Catl6) were purchased from Molecular Probes ( Eugene, O R ) . 4-0x0-2,2,6,6-tetramethylpiperidir~e-d,~ , 1 - I'N- 1 -oxyl ([15N]PDT) was purchased from MSD Isotopes (St. Louis, MO). Serum and lipo- proteins were labeled with lipid-soluble nitroxides as follows. An aliquot of nitroxides in ethanol sufficient to give final bulk concentrations of nitroxides from 0.1 to 1 .O m M was pipetted into an NMR tube and dried over nitrogen to make an uniform film of nitroxide on the side wall of the tube. The serum or lipoprotein was then added and vortexed intermittently for 30 min, and the suspension allowed to stand for 15 min. Sodium azide was added to the sample preparations to retard bacterial growth.

All NMR spectra were measured at room temperature on a Bruker AM 360 spec- trometer operating at 8.5 T. Free induction decays were collected in 8 k data points for IH, 2 k data points for 31P, and 16 k data points for 13C after 60" pulses. Delays were 2 s for 'H, 3 s for 31P, and 0.5 s for I3C. The number of scans were 200 for 'H, 800 for 31P, and 50,000 for I3C. Composite pulse IH decoupling was applied for ac- quiring 13C and 31P NMR spectra. The water proton signal near 4.63 ppm was sup- pressed by selective radiofrequency saturation except during data acquisition. Area integrals of lactate resonances were obtained using a Lorentzian line-fitting program (Glinfit, Bruker Users Society). The curve fitting was based on the spectra of lactate in solution and did not include the small peak just upfield from the lactate; we believe

122 CHEN ET AL.

that the latter probably arises in protein. The sample volumes were 0.5 ml for 'H and 2.5 ml for 31P and I3C. All electron paramagnetic resonance (EPR or ESR) spectra were obtained at room temperature on a Bruker 200D-SRC spectrometer ( TEloz cavity) with a 50-pl sample, under 5 mW power and I G modulation.

Extracts of serum were prepared by the technique of Folch ( 10). Chloroform, meth- anol, and water are present in the proportion of 8:4:3 by volume, allowing for a separation of the lipid containing chloroform/ methanol phase, and water/ methanol phase after centrifugation at 4000g for 15 min. The interface between the two liquid phases principally consists of precipitated denatured protein and is discarded.

'H and 13C NMR resonances assignments were taken from Steim el al. ( 11) and Godici and Landsberger ( 12).

For chemical analysis, lactate was converted to pyruvate with excess NAD and lactate dehydrogenase. Pyruvate formed during the reaction was trapped with hydra- zine. The increased absorbance at 340 nm due to NADH formation is a measure of lactate originally present.

RESULTS

As shown in Fig. 1, 5-doxy1 stearate suppresses the peaks of methyl of choline and the methyl and methylene peaks of lipids in the 'H NMR spectra of serum allowing the lactate peaks to stand out clearly. These methyl and methylene peaks are suppressed completely when the bulk concentration of 5-doxy1 stearate in serum is 0.75 m M or higher. To determine whether this effect is due to paramagnetism, or perturbations of

4.0 3.5 3.0 2.5 2.0 1.5 1.0 PPM

4 0 3.5 3.0 2.5 2 0 1.5 1.0 PPM

FIG. 1. Effects of 5-doxy1 stearate on the 'H NMR spectra of serum. Spectra with different bulk concen- trations of nitroxides are shown (mM).

FIG. 2. 'H NMR spectra of serum with 5-doxy1 stearate and added lactate. The amounts of lactate added are shown (mg/ 100 ml).

SELECTIVE SUPPRESSION OF LIPID PEAKS 123

the motion or structure of the system by nitroxides, we reduced the paramagnetic nitroxides to the diamagnetic hydroxylamines by ascorbate plus a nitroxide [ "N ] PDT (0.1 m M ) . The nitroxide was introduced as an electron shuttle ( $ 1 3 ) between aqueous and hydrophobic regions, because ascorbate is charged and does not dissolve in the hydrophobic region where 5 doxyl stearate is located. The reduction process was mon- itored by EPR spectroscopy (14 ) . We found that the suppressed lipid peaks could be restored upon the reduction of 5-doxy1 stearate, indicating that the observed effect was due to paramagnetism. There was no detectable reduction of the nitroxides by the serum during the experiments.

To determine if this method could provide quantitative data on the amount of lactate in serum, we added exogenous lactate to the sample. The intensity of the resonances of lactate increased linearly with the amount of lactate that was added (Figs. 2 and 3 ). We then could calculate the original concentration of lactate in serum (9.5 mg/100 ml) from the intercept of a linear plot of the areas of the doublet of lactate versus the amount of lactate added. The concentration of lactate was measured independently by the enzymatic assay as described under Materials and Methods and the value measured (9.8 -t 0.9 mg/dl, three determinations) agreed with the NMR measurement (9.5 f 0.4 mg/dl, five determinations).

The observed lipid peaks in serum arise from lipoproteins while the doxyl stearates distribute into all serum proteins, especially albumin. We, therefore, determined the amount of doxyl stearate required to suppress the lipid peaks in a solution of lipo- proteins and found that the concentration of 5-doxy1 stearate required to cause the suppression of lipid protons in lipoproteins is about an order of magnitude less than in the serum under the comparable conditions.

The amount of Catln required to suppress the spectra of the methyl and methylene protons was about an order of magnitude less than that needed for an equivalent effect by the doxyl stearates (e.g., 0.1 m M vs 1 .O m M gave equivalent effects). This is due

0.0 5.0 10.0 15.0 20.0 25.0 30.0 Added Lactate (mg/dl)

FIG. 3. Quantitative analysis of added lactate by NMR. The spectra in Fig. 2 were integrated as described under Materials and Methods.

124 CHEN ET AL

to the different affinities of the nitroxide for serum proteins; doxyl stearates bind very well to albumin at the sites at which fatty acids usually bind (15 ) while Catl6 binds poorly to albumin. The differential affinity of these nitroxides was confirmed by ex- periments in which albumin was added to serum; this decreased the effects of doxyl stearates but had little effect on the broadening effects due to Catl6.

As shown in Fig. 4, Catl6 completely broadened out the I3C peaks of the methyl of choline, while it significantly broadened the I3C peaks of the methyl and methylene of the hydrocarbon chains of phospholipids. This is consistent with the results shown in Fig. 1 where 5-doxy1 stearate suppressed the peaks of methyl and methylene in the 'H NMR spectra of serum.

The effects of the nitroxides on 31P NMR spectra of serum are shown in Fig. 5. There are three peaks in the 3*P NMR spectrum of serum. The left one was identified as inorganic phosphate by its chemical shift. The molecules responsible for the other two peaks were extracted into the chloroform/methanol phase of the extracts. The right peak has the chemical shift of phosphatidylcholine in low density lipoprotein ( 1 6 ) , and the middle rather broad peak has the chemical shift of sphingomyelin and probably also has contributions from other phospholipids (16) . Catl6 and 5-doxy1 stearate broadened the phosphatidylcholine peak more than the sphingomyelin peak. The effect of 12-doxy1 stearate on sphingomyelin was similar to that of 5-doxy1 stearate

I I I I I I I 70 60 50 40 30 20 10

PPU

FIG. 4. Effects of Cat,6 on the "C NMR spectrum of serum. Spectrum ( A ) is from a control sample, and spectrum (B) is from a sample with the addition of 0.5 mM Catl6. Cat,, broadened out the peak due to the methyl carbon of choline while it slightly broadened the peaks of methyl and methylene carbon of the hydrocarbon chain of phospholipids.

SELECTIVE SUPPRESSION OF LIPID PEAKS 125

FIG. 5. Effects of nitroxides on the 31P NMR spectra of serum; A-control, B-5-doxy1 stearate, C-12- doxy1 stearate. In the control spectrum, there is a two component peak from phospholipids located upfield from Pi( A). 5-doxy1 stearate completely broadened out the high-field component and moderately broadened the low-field component (B); Catl6 had a similar effect. The effect of 12-doxy1 stearate on the low-field component was similar to that of 5-doxy1 stearate and Cat,, but it did not broaden the high-field component as effectively as the other two nitroxides (C). The concentration of nitroxides used was 0.5 mM.

and CatI6; it did not broaden the 31P peak due to phosphatidylcholine as effectively as the other two nitroxides.

These results raise some interesting questions about the precise location of the ni- troxides in the lipoproteins. If, as seems most likely, the effects are due to dipole- dipole broadening or other effects that have the usual 1 / r 6 dependency, the effects of each nitroxide should be very sensitive to the location of the nitroxide group. The results for 5- and 12-doxy1 stearates, especially in regard to their effects on the 31P spectra fit reasonably well with such an assumption, because these are expected to have an average location 5 or 12 carbons deep from the surface of the organized lipids and to have a moderate amount of movement perpendicular to the surface of the lipids. The nitroxide group of Catl6, however, is expected to be near the surface of the lipids and its efficacy in suppressing the methylene resonances is surprising and seems to indicate that it has significant movement deep into the hydrocarbon region of the organized lipids.

While the use of nitroxides to affect NMR spectra and thereby probe bimolecular structures has been emphasized previously by others, ( 17-20), we believe that our results add some significant new aspects. In addition to the directly new application of facilitating the measurement of lactate in serum, we suggest that these results indicate the feasibility of this approach to enhance the information available from NMR spec- troscopy of complex biological systems, including intact animals and perhaps patients.

126 CHEN ET AL.

The use of appropriate contrast agents that would localize in specific areas because of their physical-chemical properties (e.g., lipid solubility, charge, binding to specific receptors) might be very useful for enhancing the quality of the information available from NMR spectroscopy of cells, tissues, and animals. This approach might make it feasible to determine the contributions of different chemical species to an observed NMR peak, such as the various phosphorus esters, by using contrast agents which have their paramagnetic groups at different depths in membranes. Metabolically re- sponsive contrast agents such as the nitroxides might enable one to distinguish the overlapping contributions of species located in cell compartments with different met- abolic activities; e.g., the rate of reduction (and hence loss of peak suppression) of nitroxides in the membranes of mitochondria would be expected to be greater than the rate of reduction of nitroxides in nuclear membranes.

DISCUSSION

These results demonstrate the feasibility and utility of using lipid soluble nitroxides to obtain selective effects on NMR spectra. Using 'H spectroscopy the resonances due to methyl and methylene protons were suppressed sufficiently by doxyl stearates to allow the resolution of the underlying lactate peaks and quantitative determinations of the amount of lactate, and this quantitative determination was confirmed by an enzymatic assay. Similarly selective effects on methyl and methylene carbon spectra were found with I3C NMR studies.

It was expected that Catl6 and 5-doxy1 stearate would affect the phosphorus spectra of the phospholipids and not the inorganic phosphate because the latter is distributed throughout the sample while the Catl6 or 5-doxy1 stearate is localized in the lipoproteins. The differential effects of Catl6 or 5-doxy1 stearate on the two phosphorus peaks of phospholipids indicates that these peaks are from phospholipids which may have sig- nificant differences in their sites in the lipoproteins. This could be the result of different ratios of the phospholipids in different lipoproteins which have different partition coefficients for Catl6 and 5-doxy1 stearate and/or different locations of the phospho- lipids in the same lipoproteins. The nitroxide group on 5-doxy1 stearate is localized at the fifth carbon from the carboxyl group on the hydrocarbon chain, whereas it is at the twelfth carbon for 12-doxy1 stearate. It should be pointed out that when dealing with lipophilic compounds in heterogeneous systems the local concentrations of the lipophilic compound may be very high and cannot be calculated directly from the bulk concentration. One needs knowledge of the amount of lipid that is present and the partition coefficients for each type of lipid. In organized lipid systems such as bilayers, the partition coefficients can change considerably when there is a change in structure (e.g., change of phase).

Overall these studies indicate that the use of lipophilic nitroxides as contrast agents for MRS of biological systems may be productive. Significant effects on 'H, I3C, and 31P spectra were obtained with the addition of relatively low amounts of nitroxides. Even in this relatively simple system (simple compared to cells or tissues) interesting and potentially useful differences of localization in different lipids were found. Finally, the specific application of using the nitroxides to obtain quantitative estimates of lactate may be of immediate use for studies of metabolism. The use of Catl6 for this purpose seems especially promising because unlike the doxyl stearates which bind to

SELECTIVE SUPPRESSION OF LIPID PEAKS 127

albumin, it localizes preferentially in lipoproteins and therefore was effective at quite low doses.

ACKNOWLEDGMENTS

The authors gratefully acknowledge Dr. Shi-Jiang Li for helpful discussions, Dr. David G. Lange for the use of the ESR spectrometer, and Dr. V. P. Chacko for extensive advice and assistance with the NMR measurements. This work was supported by grants from the American Cancer Society (K.C. and H.M.S.) and by Grants CA 44703 (J.D.G.) and CA 48266 (J.P.W.) from the National Institute of Health.

REFERENCES

1. H. M. SWARTZ, in “Advances in Magnetic Resonance Imaging” (E. Feig, Ed.), p. 49, Ablex Publishing,

2. K. CHEN AND H. M. SWARTZ, Biochim. Biophys. Acta 970,270 ( 1988). 3. K. CHEN AND H. M. SWARTZ, Biochim. Biophys. Acta 992, 131 (1989). 4. H. M. SWARTZ, K. CHEN, M. PALS, M. SENTJURC, AND P. D. MORSE, 11, Magn. Reson. Med. 3, 169

5. K. CHEN, J. F. GLOCKNER, P. D. MORSE, 11, AND H. M. SWARTZ, Biochemistry 28, 2496 ( 1989). 6. A. DANIELS, R. J. P. WILLIAMS, AND P. E. WRIGHT, Nature (London) 261, 321 ( 1976). 7. D. L. RABENSTEIN AND T. T. NAKASHIMA, Anal. Chem. 51, 1465A (1979). 8. J. K. NICHOLISON, M. P. OFLYNN, AND P. J. SADLER, Biochem. J. 217, 365 ( 1984). 9. V. N. SCHUMAKER AND D. L. PUPPIONE, in “Methods in Enzymology” (J. P. Segrest and J. J. Albers,

Nonvood, NJ, 1989.

(1986).

Eds.), Vol. 128, p. 155, Academic Press. San Diego. 1986. 10. J. FOLCH. M. LEES, AND G. H. STANLEY, J. Biol. Chem. 226,497 ( 1957). 11. J. M. STEIM, 0. J. EDNER, AND F. G. BARGOOT, Science 162,909 ( 1968). 12. P. E. GODICI AND F. R. LANDSBERGER, Biochemistry 13, 362 ( 1974). 13. D. 0. NETTLETON, P. D. MORSE, 11. AND H. M. SWARTZ, Arch. Biochein. Biophys. 271,414 ( 1989). 14. K. CHEN. P. D. MORSE, 11, AND H. M. SWARTZ, Biochim. Biophys. Acta 943, 477 ( 1988). 15. J. D. MORRISETT, H. J. POWNALL, AND A. M. GOTTO, JR., J. Biol. Chem. 250,2487 ( 1975). 16. P. L. YEAGLE, R. B. MARTIN, L. POTTENGER. AND R. G. LANGDON, Biochemistry 17,2702 (1978). 17. P. A. KOSEN, in “Methods in Enzymology” (N. J. Oppenheimer and T. L. James); Vol. 177. p. 86,

18. J. ANGLISTER, in “Biological Magnetic Resonance,” (L. J. Berliner and J. Reuben, Eds.), Vol. 8, pp.

19. S. WEINSTEIN. B. A. WALLACE, E. R. BLOUT, J. S. MORROW, AND W. VEATCH, Proc. Nail. Acad. Sci.

20. P. MEERS AND G . W. FEIGENSON, Biochim. Biophys. Acta 938,469 ( 1988).

Academic Press, San Diego, 1989.

597-614, Plenum, New York, 1989.

USA 76,4230 (1979).