1997;57:407-414. Published online February 1, 1997.Cancer Res Chantal Rémy, Nathalie Fouilhé, Ignasi Barba, et al. H Nuclear Magnetic Resonance Correspond to Lipid Droplets

1Evidence That Mobile Lipids Detected in Rat Brain Glioma by   

  

Updated Version http://cancerres.aacrjournals.org/content/57/3/407

Access the most recent version of this article at:

  

Citing Articles http://cancerres.aacrjournals.org/content/57/3/407#related-urls

This article has been cited by 8 HighWire-hosted articles. Access the articles at:

  

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

SubscriptionsReprints and

.pubs@aacr.orgDepartment atTo order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions.permissions@aacr.orgDepartment at

To request permission to re-use all or part of this article, contact the AACR Publications

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

[CANCERRESEARCH57,407—414,February1, 1997]

ABSTRACT

Mobile lipids have been detected by proton nuclear magnetic resonance(NMR) Inanimal and human tumors (cultured cells, biopsies, and in vivo),but their origin and subceilular location are still unclear. They have beenassociated with malignancy, metastatic abifity, drug resistance, and necrosis. We wanted to determine whether these lipids are located withinplasma membrane microdomains or In lipid droplets for a C6 cell-inducedrat glioma.NMR-visiblemobilelipids were found in all subcellular fractions Isolated from the rat tumor, except In the cytosolic supernatants.Transmission electron microscopy showed that lipid droplets were presentin all subcellular fractions containing NMR-visible lipids and In thenecrotic and perinecrotic areas of the tumor. The mean diameter ofdroplets Isolated by flotation in the subcellular fractionation protocol was0.97 pm (n = 682; droplet profile diameter range between 0.2 and 5.0,.tm).The apparent diffusion coefficient for these lipids (46 ±17 pm2s')measured in vivo by proton spectroscopy was four orders of magnitudehigher than would be expectedif mobile lipids were Insideplasma membraise microdomahas. The combined results demonstrated that mobilelipids detected in vivo by proton NMR in the C6 rat glioma are located inlarge lipid droplets, associated with the necrotic process.

INTRODUCTION

Since 1982, numerous studies have described the presence of anarrow and intense resonance in the 1.3-ppm region of proton NMR@spectra obtained from cultured cells (tumoral, embryonic, and stimulated cells), from experimental solid tumors (rat mammary adenocarcinomas), and from human tumor biopsies (ovarian and colorectal;Refs. 1—5).This resonance was not detected in the proton spectra ofthe normal cells and tissues and was considered to be a potentialmarker of malignancy (2). This resonance was assigned to the(CH2)n protons of lipids located in the cellular plasma membranes

(1). However, the relatively long spin-spin T2relaxation time of theselipids (value of the order of 200 ms; Ref. 2) indicated that they werehighly mobile and thus unlikely to be organized in a bilayer structure.It was proposed, based on a variety of experimental data (reviewed inRefs. 2 and 3), that the mobile lipid signal detected by ‘HMRS intumoral and stimulated cells originates from neutral lipids (mainlytriglycerides) arranged in microdomains (25—28nm in diameter)embedded within the phospholipid bilayer of the plasma membrane.The MRS detection of these neutral lipid microdomains was found tobe correlated with cellular proliferation, cellular stimulation, acquisition of resistance to anticancer drugs, and metastasis (2, 3). The

Received 8/5/96; acce@ 12/4/96.The costs of publication of this article were defrayed in part by the payment of page

charges. This article must therefore be hereby marked advertisement in accordance with18 U.S.C. Section 1734 solely to indicate this fact.

I This work was supported by the Association Espoir, l'Is@re contre le Cancer,

Fondation pour Ia Recherche Médicale,and by Comisiôn Interministerial de Ciencia yTecnologIa (CICYT@SAF 93-0582-C02-Ol. Short-term travel grants were obtained fromCooperation franco.espaguole-Accords INSERM-CSIC. N. F. received a grant from theLigue Nationale Contre Ic Cancer, and E. S-L. received a grant from the Association pourla Recherche surle Cancer and Ligue Nationale Contre le Cancer. A. M. held a Fellowshipfrom the AsociaciônEspaflola Contra el Cancer during his participation in the study.

2 To whom requests for reprints should be addressed.

3 The abbreviations used are: NMR, nuclear magnetic resonance; MRS. magnetic

resonance spectroscopy.

mobile lipid resonance of metastatic malignant cells and tumor biopsies with a high metastatic potential was found to be characterized by

the presence of a component with a very long T2 (between 500 and1000 ms), giving rise on two-dimensional correlation spectroscopyspectra to a specific cross-peak (1.3—4.0ppm; Refs. 2 and 3). Thiscomponent was assigned to fucosylated species located on the surfaceof the membrane of the metastatic cells (5). However, it was not clearwhether these species were part of the neutral lipid microdomains (3).

In 1993, Callies et a! (6) showed that the presence of the mobilelipid resonance in the proton NMR spectra of myeloma cells, grownin the presence of oleic acid, was correlated with the formation ofcytoplasmic lipid droplets. While observing lipid droplets in their cellpreparations, Mackinnon et al. (4) and May et al. (7) interpreted theirdata as showing that the mobile lipid resonance detected by ‘HMRSoriginates mainly from the plasma membrane and not from intracellular lipid droplets.

In previous in vivo ‘HNMR studies, we showed that an intracerebral C6-induced glioma presented significant spectral differencescompared to normal brain tissue (8, 9). One of the most remarkabledifferences was the presence, even at long echo times (272 ms), of anarrow and intense resonance in the 1.3-ppm region of the spectra. Invivo and ex vivo one-dimensional editing, two-dimensional correlationspectroscopy, and two-dimensional f-resolved experiments indicatedthat this resonance was due to the overlapping of the lactate resonancewith those of the (CH2)n protons of mobile lipids (10, 11). Thismobile lipid resonance was also detected by other groups in ‘Hspectra of human brain tumor biopsies (12—14)and in vivo (15—17).Most of these tumors were high-grade glioma and metastasis. Kueselet a!. (12, 13) showed that the amount of mobile lipids detected by ‘HMRS correlated with the amount of necrosis in the brain tumorbiopsies. However, the subcellular location of these mobile lipids(cytoplasmic lipid droplets or neutral lipid microdomains within themembranes) were unknown (12, 13).

The aim of the present study was to defme the location of themobile lipids detected in vivo by ‘HMRS in the C6 glioma. For thispurpose, in vitro and in vivo experiments were performed. The in vitroexperiment consisted of subcellular fractionation. The excised tumorand subcellular fractions were analyzed by proton NMR spectroscopyand transmission electron microscopy. The in vivo experiment consisted in NMR measurement of the mobile lipid diffusion characteristics (for a review of the technique, see Ref. 18). This experimentgives insight into the physico-chemical properties of the mobile lipidcompartment, i.e., whether the NMR-visible lipids are confined insmall membrane microdomains (25—28am in diameter) or in dropletsor vesicles of larger size.

MATERIALS AND METHODS

Experimental Model oflntracerebral Glioma. Intracerebral tumors wereinduced by C6 cells, a cell line derived from a chemically induced malignantrat glioma (19). Ten thousand C6 cells suspended in 10 @dof a 1% agarsolution (jweparedin DMEM supplemented with 10% FCS, 4 nmi glutamine,50 units/mi penicillin, and 50 pg/mi streptomycin) were stereotactically injected 3 mm under the dura in the right caudate nucleus of female Sprague

407

Evidence That Mobile Lipids Detected in Rat Brain Glioma by 1H NuclearMagnetic Resonance Correspond to Lipid Droplets'

Chantal Rémy,Nathalie Fouilhé,Ignasi Barba, Ernest Sam-Lal, Hana Lahrech, Maria-Gracia Cucurella,Marguerite Izquierdo, Angel Moreno, Anne Ziegler, Raphael Massarelli, Michel Décorps,and Carles Ants2INSERM U438, RMN Bioclinique, CHU de Grenoble, BP 217, 38043 Grenoble Cedex 9, France (C. R., N. F., E. S-L, H. L, M. L, A. Z, R. M., M. DI, and Universitat Autonômade Barcelona, Departament tie Bioqu(mica i Biologia Molecular, Facultat de Ciències, 08193 Bellaterra, Spain (I. B., M-G. C., A. M., C. A.)

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

GUOMA LIPID DROPLEtS: NMR AND MICROSCOPY STUDY

Dawley rats (20). Survivaltime is around4 weeks. The rats were sacrificedbetween 21 and 26 days after cell implantation with an overdose of anesthetic(400 mg/kg chloral hydrate), and the brain was excised. The tumor and the leftnormal hemisphere were separately dissected on ice. All operative proceduresand animal care strictly conformed to the Guidelines of the French Government (decree no. 87-848 of October 19, 1987, licenses 006722 and A3807l,respectively).

Subcellular Fractionation Protocol. The subcellular fractionation procedare (Fig. 1) was a modification of protocols described for mammalian brain(21, 22). The tissue used for fractionation was the whole brain (exceptcerebellum) for control animals and the contralateral hemisphere (exceptcerebellum) or the whole tumor for tumor-bearing animals. The excised tissuewas weighed and homogenized with a glasslTeflon poner (15 strokes at 2000rpm speed) in sucrose buffer 1 [0.32 M sucrose, 1 mM EDTA, and 30 ms@sodium phosphate buffer (pH 7.4); 10mug wet weight]. The homogenate (lIT)was centrifuged at low speed (270 X g for 20 rain). The crude pellet (P1) wasresuspendedin a sucrosebuffer2 [2.4 Msucrose,10 mMsodiumphosphatebuffer, and 1 mr@iMgCl2 (pH 7.4); 10 mUginitial wet weight] and centrifugedat 54,800 x g for 40 mm to obtainanenrichednuclearfraction(P2) anda clot(SL2) floating over a clear supernatant(S2). The crudesupernatant(Si) wascentrifuged (103,800 X g for 2 h 30 rain). The pellet thus obtained (P3) wouldcontain mitochondria, microsomes, and plasma membranes. In the case oftumoral tissue, a lipid film (SL3) was obtained, floating over a clear superna

tant (S3). During the course of two subcellular fractionations, a part of the

crude supernatant (51) was directly centrifuged at 103,800 X g for a longertime (12 h) to achieve a better separation of the lipid film, which was namedSM from the pellet P4 and clear supematantS4. The entire protocol wasperfonned at 4°C.The volume of each fraction was measured. Aliquots of eachfraction were collected for NMR examination and electron microscopy.

Seven separate subcellular fractionation experiments were performed: sixfor the tumortissue (pool of five to nine tumorsin each case; total weight,1.0—2.7g between 21 and 26 days after cell implantation) and one for thecontralateral hemisphere (pool of two contralateral hemispheres; total weight,1.4 g). Some modifiedfractionationprotocolswere appliedthree times onnormal rat brains and three times on tumor contralateral parenchyma. Mobilelipids were never detected in the proton spectrum obtained from any of thesubcellular fractions isolated from normal or contralateral rat brain.

NMR spectroscopy was performed on the subcellular fractions obtainedfrom the seven standard experiments and electron microscopy analysis onthose obtained from three of them (two tumoral tissues and the contralateralhemisphere). No other homogeneity test, apart from electron microscopyexamination,was perfonned.

Electron Microscopy. Electron microscopy was carried out for both tumoral and contralateral tissues and for the following subcellular fractions: HT,P1, Si, P2, SL2, P3, SL3, P4, and SM. Electronmicroscopyexaminationofone tumor was perfonnedon three contiguouscubes, about 1 mm@,takenserially from the superficial to the central part of the tumoral tissue. Eachsample was fixed overnight at 4°Cwith a 2.5% glutaraldehyde solution incacodylate buffer (0.1 M,pH 7.4) and postfixed with 1% osmium tetroxide incacodylate buffer (0.1 M, pH 7.4) for 12 h at 4°C.After each fixation, sampleswere washedfourtimes withcacodylatebuffer(0.1 M,pH 7.4) andkeptin thesa,ne buffer at 4°Cuntil further processing. To avoid loss of lipids due tofixationproblems(23), floating lipid layers SL2, 5L3, and SM were directlyfixed with 1%osmium tetroxide in cacodylate buffer and washed four times asabove.

Fixed samples were taken to the “Serveide Microscopia Electronics―(UniversitatAutonömado Barcelona)for furtherprocessing.Samplesweredehydrated and embedded in Spurr's resin containing uranyl acetate. Ultra-thinsections (50-70 am) ofthe samples were stained with lead citrate. Observationwas made with an Hitachi H-7000 transmission electron microscope.

The mean lipid droplet diameter(I) was calculated for the external andinternal part of tumor and for two SL2 fractions, according to classicalstereological techniques (24, 25). For each sample, the ultra-thin sectionanalyzed was chosen randomly. For the 5L2 fractions, all droplet profiles ofthe section were digitiZedusing a Bio-Rad GS-700 imaging densitometer(Bio-Rad, Madrid, Spain). In the case of the excised tumor, only a smallnumber of droplets were digitiZed. The diameter of the droplet profiles wasmeasured on the digitized micrographs using the LEICA Q500MC software(Cambridge,Ltd., Cambridge,United Kingdom),assumingthat the shape ofthe dropletswas spherical.A very small proportionof dropletprofiles withaxial ratios larger than 1.5 were rejected. The mean lipid droplet diameter (J)wascalculatedfromthemeasureddiameter@ of eachdropletprofileusingtheformula of Fullman for ungrouped profile diameter data (25):

- IT N

whereN is the total numberof dropletprofiles measured.In Vitro NMR Experiment. Aliquots (about 350 pA)of lIT, 51, 513, S2,

and 53 fractionswere analyzedby ‘HMRS withoutany furtherdilution.TheSL2, P2, and P3 fractionswere resuspendedin sucrose buffer 1 and the Ptfraction in sucrose buffer 2. In vitro NMR experiments were performed at 37°C

(A)21 1 1 1

—+—+—+...+—

4,, 4@24@3 4N

Supematant SiFig. 1. Schema of the subcellular fractionation procedure. Sucrose

buffer 1 was a 30 mi@sodium phosphate buffer containing 0.32 Msucrose and 1mMEDTA(pH 7.4). Sucrose buffer 2 was a 10 inst sodiumphos@ buffer containing 2.4 Msucrose and 1 ms@MgCI2 (pH 7.4).

408

TI sue

1@ + Sucrose buffer 1

HomogenateHI

PelletPi

+ Sucrose buffer 2F—10ml/gwetweight@ g 40 mm 4°a)

SL2 Clot

S 2 Hydrosolubl•traction

P2 Nuciar fraction

0o3800g 2 h 30 mm 4°c)

Llpidfiim@ 5L3

Cytoaol S 3

Mitochondria,@ 3mlcrosom•s,

plasma mimbranac

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

(B)

vwas set IL)184 @ts,which gave an excitation maximum at 1.3 ppm. The echotime (TE) was 2 ma, and the repetition time (TR) was 1.2 s. Depending on thedilution of the sample, 512 or 1024 scans were accumulated.Data wereacquired over a 5,000-Hz sweep width, digitized with 4,0% or 8,192 datapoints and zero-filled to 16,384before Fourier transformation. For each sampleobtained from an individual subcellular fractionation protocol, the same capillary (1.1-mm outer diameter) of sodium 3-trimethylsilyl [2,2,3,3@2}I4]propionate (20 mM in D20) was used as an external standard for chemical shift andquantitation purposes. The area of the mobile lipid resonance in each fractionwas normalized against the area of the sodium 3-trimethylsilyl [2,2,3,3-2H@,]propionate resonance and expressed, after correction for dilution of the variousfractions, with respect to those of the initial homogenate.

In Vivo NMR Experiments and Principles. In vivo diffusion measuremania were carried out using a 4.7 T vertical bore Imaging/Spectroscopy

409

Table 1 Quantification ofthe NMR-visiblemobile lipid resonance area at 1.26ppmrecoveredin each fraction, defined as a percentage of thehomogenate°P1

29.9±17.9 Si 52.8±14.3SL212.9±5.9 SL3 17.4±7.1S2

ND 53NDP219.6±13.5 P37.1±4.8a

Mma values ±SD; n = 6 separate experiments. ND, not detectable.

OUOMALIPIDDROPLFfS:NMRANDMICROSCOPYSTUDY

system(BrokerCXP)equippedwithanunshieldedgradientsystembuiltin ourlaboratory. A 15-mm diameter circular surface coil was screwed onto the skullof the animalas describedpreviously(27). The rats(n = 6) were anesthetizedwith a mixture ofhalothane and air(2% to establish the anesthesia and between0.4 and 0.8% duringthe NMR measurements).Body temperaturewas controtted and maintained at 37°Cwith a water-based heating blanket. Spatiallylocalized diffusion measurements were achieved using a stimulated-echo sequence (28, 29). The position of the volume of interestwas defined usingmultislice coronal and transverse pilot scans (TRITE,2000/80 ma) either in thetumoror in thecontralateralhemisphere.The echo timeof the stimulatedechosequence was TE 136 ma and the repetition time TR = 2 s. A mixing timeof 52 mawas chosen to minimizethe signal of lactate(30). Diffusion weighting was achievedusing gradientpulses oflength 8 = 21 mawith @t= 86 mabetween rising edges, leading to a diffusion time @dif=@@ 8/3 = 79 ma. Thediffusion-encoding gradient was parallel to the main field B0. The gradientfactor b is defined as follows (31):

b = ‘y@G282(@—8/3) (C)

whereG is the gradientstrengthand‘ythe protongyromagneticratio.Foreachanimal, 10 spectra were recorded with gradient factors b ranging between 400and 10,000 @@-2

To determinethe diffusioncoefficientD, the resonanceline areaof mobilelipids at 1.26 ppm, A(b), was fitted to Eq. D using a least-square method:

A (b) = A (0) . exp(—b. D)

SL2

(D)

If one can assume that diffusion is not hindered by boundaries (unrestricteddiffusion), the distance over which the molecules move during the diffusiontime l@5jfused in the experiment (79 ma) can be expressed as the root meansquare displacement, A, of the mobile lipids in the direction of the appliedgradient.This is given by:

A = @j2. D . (E)

Thus, for free diffusion, A increases with diffusion time. When restricteddiffusion occurs, A reaches a limit that depends on the compartment size, andthe diffusion coefficient calculated from Eq. D is known as the apparentdiffusion coefficient (ADC; Ref. 32). In any cases, the root mean squaredisplacement calculated with Eq. E, using the measured ADC, yields anestimateof the minimumsize of the compartment.

Diffusion measurements are inherently sensitive to motion because the pair

of gradient pulses sensitizes the NMR signal for displacementsof a fewmicrometers. Besides random diffusion of lipid molecules, the observed decrease of the signal at the large b values used in this experiment could be dueeither to physiological movements or to mechanical vibrations of the sampleafter the application of strong gradients. To check that this was not the case,one animal was sacrificed inside the magnet by an overdose of halothane.Then, the lipid diffusion coefficient measurement was repeated immediatelypostmortem. No appreciable difference was found with in vivo measurements,ruling out the hypothesis ofa decrease ofthe lipid signal as a function ofb dueto physiological movements. In addition, no appreciable signal decrease wasdetected when measuring diffusion a high molecular mass polymer (polydimethylsioxane) within the same b range. This ruled out the presence ofmechanical vibrations caused by eddy currents. Thus, we concluded that theroot mean square displacements reported in this study were not influencedeither by physiological motions or by mechanical vibrations of the gradient!sample system.

I.I.,,-,—@, I

1.8 1.6 1.4 1.2 1.0Pr,,

Fig. 2. Expanded 1.3-ppm region ofpeoton spectra ofthe subcellular fractions isolatedfrom rat gliomas.The spectrawere plottedat differentmagnificationfor qualitativecomparison.For quantitativeresults,see Table I. Arrow,positionof the mobilelipidresonance. A correction for a frequency shift induced by high sucrose concentration (2.4M) was made before plotting.

with a 9.4 T magnet Broker AM400. The pulse sequence combined waterpresaturation (800 ma at 6 mW) with selective excitation using a phase

gradient free spin-echo sequence based on the use of Jump-and-Return andbinomial pulses (26):

90°—i@—90°—TE/2 —90°—2r —90°—TE/2 —acquisition

.8 :6

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

GUOMA LIPID DROPLSTS: NMR AND MICROSCOPY STUDY

.1*

@;1. @?‘@p@

@ @r@

4...@. @:

.-@@&:,. S

.4'

@%

t:@:1@

@‘%@@•a.@ .•@@@ .-‘

@‘‘@:!@:@,@_‘)@.t@j:s

@-@

@:

@tS

•@@@:@‘1•@

(...

S@-@

Fig.3.Transmissionelectronmicrographsof glioma.a, externalpartof thetumor.Cellsarewelldelimited;mitochondria,endoplasmicreticulum,andlipiddroplets(arrows)canbeobservedin thecytoplasm.b, enlargementof thesquaredareaina. Somelipiddropletsappearto besurroundedbya membrane(arrows).c—c,internalpartof thetumor.In c, cellsarenotwelldelimited;cellulardebrisandlipiddroplets(arrows)canbeobserved.Ind ande, dropletsor dropletsarrangedin “grape―shapeareshownat a highermagnification.Bar,l@ni

doublet of lactate at 1.3 ppm and the mobile lipid resonance at 1.26ppm (Fig. 2). This mobile lipid resonance was detected in all fractionsobtained from the tumoral tissue [P1, 51, P2, SL2, P3, and SL3 (Fig.2) and P4 and SM (data not shown)] with the exception of thesupernatants 52, 53, and 54. The quantitative data (Table 1) indicatethat 83% of the mobile lipids detected in the homogenate spectrum

410

a@. @.t

@@:‘@:

@@ .@

. ‘

@;:-@

RESULTS

Proton NMR Spectroscopy of Subcellular Fractions. The 1.3-ppm region of proton spectra of the homogenate and subcellularfractions (data not shown) isolated from contralateral hemispheresshowed no mobile lipid resonance in any of the spectra. The samespectral region for the tumor homogenate was characterized by the

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

GLIOMALIPIDDROPLETS:NMRANDMICROSCOPYSTUDY

S •,@ _@

@.

S@ @‘@@@ @. .@, r@

@ ‘@ :‘@ “

by a membrane (Fig. 3, a and b). From the surface to the center of thetumoral tissue, the electron microscopic observation revealed an increasing number of cells showing ill-defmed plasma membrane andendoplasmic reticulum, a decreasing number of mitochondria, and anincreasing number of vacuoles. In addition, lipid droplets were foundto increase in number and in size toward the clearly necrotic core ofthe tumor (Fig. 3, c—e).The mean lipid droplet diameter 1 was foundto be 0.40 p.m (range of measured diameters of droplet profiles, 4,

found in this area. Some of these droplets appeared to be surrounded between 0.05 @mand 1.6 p.m; n = 93) in the external part of the411

•?@@ ,,

J@.

4'

‘@•:@@

t# ::@t@.

1.@@@@

45 5

SS'@@SS

‘@@ S

._@SS ; S.@

@a*.@S;J@ @..@@ ;@.S S.d

Ftg. 4. Transmission electron micrographs of subcellular fractions isolated from rat gliomas. a, P2 fraction enriched in nuclei. b, SL2 fraction containing mainly lipid droplets andcellular debris. c, PS fraction enriched in mitochondria, microsomes, and plasma membranes. d and e. SM and P4 fractions (obtained from Sl fraction as P3 and SL3 after a 12-hcentrifugation time). Droplets were present in all fractions examined except in the P4 fraction, in which they were rare. Arrows in a and c point to droplets. Bar, 1 ,.un.

were recovered after the first centrifugation: about 30% in the pellet(P1) and 53% in the supernatant (Si). Only one-half of the mobilelipids contained in the homogenate were recovered after both ultracentrifugations: 19.6 and 7.1% in the P2 and P3 pellets and 12.9 and17.4% in the floating layers SL2 and SL3.

Electron Microscopic Examination of Excised Tumors. The cxtemal part of the tumor showed hyperactive cells, rich in mitochondriaand well-developed endoplasmic reticulum. Few lipid droplets were

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

GLIOMALIPID DROPLETS:NMR AND MICROSCOPYSTUDY

tumor and 1.08 @m(4, range between 0.2 @mand 1.7 @tm;n = 59)in the internal part of the tumor.

Electron Microscopy of Subcellular Fractions. No lipid bodies,either droplets or vesicles, were detected by electron microscopy ofthe fractions obtained from the contralateral hemisphere (results notshown). The P2 fraction was enriched in nuclei, and P3 containedmitochondria, microsomes, and myelin. Myelin and debris were themain constituents of the SL2 fraction.

Lipid droplets were detected in Pt, 51, P2, SL2, P3, P4, and SM

fractions obtained from tumoral tissue (Fig. 4) but not in the corresponding fractions from coniralateral tissue. Because of fixation problems, SL3 samples werelost, but SM is essentially equivalent to SL3.

The mean diameter (I) for lipid droplets present in the two SL2fractions analyzed was 0.97 @m(n = 682), with a range of measuredprofile diameters (4)) from 0.2 to 5 @m.This relatively large range isquite probably due to the polydisperse nature of the lipid dropletpopulation as illustrated in the shape of the histogram of the measuredprofile diameter classes (Fig. 5).

In Vivo Diffusion Experiments. Fig. 6 shows the decrease of thelipid peak intensity as a function of the b gradient factor, observed for

one animal. For the lipid resonance, the apparent diffusion coefficient,ADC, in the directionof the gradientwasbetween30 and74 @tm2s‘,yielding ADC 46 ±17 @m2s‘(mean ±SD, n = 6). From theseADC values,theroot meansquaredisplacementAof the mobile lipidswas calculated for each animal. This yields A 2.6 ±0.5 @m(mean ±SD, n = 6). To control the accuracy of the measurements,the ADC of the N-acetyl-aspartate resonance (2.0 ppm) in the contralateral hemisphere was measured. The value obtained 290 ±100

@m2sâ€(̃n = 4) was in reasonable agreement with results publishedpreviously of 270 ±40 @m2s‘(n = 6; Ref. 29), 160 ±10 @.tm2s(n 8; Ref. 18), and 220 ±20 @.tm2s' (n = 8; Ref. 33).

DISCUSSION

In the present study, proton NMR analysis was performed on allfractions obtained after subcellular fractionation, and quantitative datawere obtained. The results showed that mobile lipid resonance was

-0.1

0 -0.2

.0

C

-0.4

-0.50 1000 2000 3000 4000 5000 6000 7000 8000

b (s.mm2)

Fig. 6. Mobile lipid signal intensity attenuation, observed for one animal. The plotshows the logarithm of signal attenuation as a function of gradient factor b. The ADC isdeduced from a linear least square fit. In this case, ADC was equal to 56 @m2s‘,and theregression coefficient R was 0.95.

absent from the spectra of all fractions obtained from the contralateralhemisphere and from those of the hydrosoluble S2 and 53 fractionsobtained from tumoral tissue, hence confirming the hydrophobicnature of the 1.26-ppm resonance. Except these two fractions, NMRvisible mobile lipids were present in all fractions (Pt , S 1, P2, SL2, P3,SL3, P4, and SL4) obtained from the C6-induced glioma. Electronmicroscopy revealed the presence of lipid droplets of varying diameters in all these fractions. Lipid droplets were already present in theexcised tumor tissue as demonstrated by electron microscopy analysis(Fig. 3). These results suggest that lipid droplets contribute, at least tosome extent, to the resonance of mobile lipids detected in the protonspectra of C6 glioma in vivo.

Quantitative analysis shows an important loss (43%) of mobilelipids during the subcellular fractionation protocol (17% after the firstcentrifugation and again 26% after the second centrifugations). It maybe explained by the presence of sticky lipid droplets. A slightly higherproportion (with respect to the homogenate) of mobile lipids wasfound in the droplet-enriched SL2 and 5L3 fractions (30.1%) than inthe nuclear-rich P2 fraction and the microsome- and mitochondriarich P3 fractions taken together (26.7%). However, the percentage ofthe resonance peak found in SL2 and especially SL3 was probablyunderestimated because of the extreme difficulty in collecting thesefloating films. In addition, some lipid droplets were found inside othersubcellular structures, like nuclei and endoplasmic reticulum. Thiswould also contribute to the incomplete recovery of mobile lipids inthe SL2 and SL3 fractions. This may be an artifact of the subcellularfractionation protocol or may originate from a real phenomenon invivo. Large lipid droplets (up to 20 @min diameter) were found inpreneoplastic rat liver, sometimes migrating through the lumen of theendoplasmic reticulum to the nucleus (34, 35). The mitochondria- andmicrosome-rich P3 fraction showed, by electron microscopy, somecontaminating droplets, which might explain the 7.1% part of thehomogenate mobile lipid signal. However, the percentage of themobile lipid resonance was higher in SL3 (17.4%) than in P3 (7.1%),also suggesting a correlation between lipid droplets and the mobilelipid resonance.

In vivo diffusion measurements provide decisive evidence that themobile lipid resonance detected in vivo by ‘HMRS in the C6 gliomais due to lipid droplets. Small neutral lipid microdomains within the

plasma membrane could account for the long T2 of the 1.26-ppmresonance if they undergo isotropic rotational motion (3). However,the root mean square displacement (A 2.6 @.tm)of the mobile lipids,calculated from the measured ADC, greatly exceeds the size 1 (1,0.025—0.028 @m)of these microdomains (3). The small size of the

412

1 2 3 4 5 6 7 8 9 10111213 141516 17181920212223

120

100

U,

-@ 800I-

Q_ 60,.-0

c@ 40z

20

0

Class ProfileFig. 5. Distribution of the profile diameters of the lipid droplets present in the two SL2

subcellular fractions examined. Classes 1 to 13 contain measured profile diameters inranges increasing stepwise by 0.l-@m each time (i.e.. class I contains measured profilediameters smaller than 0.25 pm, whereas class 2 contains profiles between 0.25 and 0.35sm). In classes 14—16,the range increases by 0.2 pm each time, whereas for the classes17—23,the range increases by 0.5 @meach time (i.e., class 23 contains measured profilediameters between 5 and 5.5 pun).

Droplet Profile Histogram

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

GLIOMAUPID DROPLETS:NMR AND MICROSCOPYSTUDY

microdomains would indeed introduce a severe restriction to thetranslational displacement of the lipids inside the microdomains. TheADC, which would be obtained underour experimental conditions formolecules confined within such microdomains, can be estimated byusing Eq. E: ADC A2/2td,f, and replacing A by 1. The value foundwas about 4.5 iO@ @im2s ‘,which is several orders of magnitudelower than the experimental result (46 ±17 @m2s‘).Another kindof diffusion could be observed: the diffusion of the microdomainsthemselves. However, the mass of the microdomains, as well as thelimited membrane fluidity, would not allow root mean square displacements as large as 2.6 @mduring a diffusion time of about 79 ms.

Thus, one can conclude that the mobile lipids observed by ‘HMRSare contained in large size compartments that appear to be lipiddroplets. However, the root mean square displacement, A = 2.6 @.tm,determined from in vivo diffusion measurements, exceeds the meandiameter of the droplet of CI)= 0.97 @.tm,which was determined fromelectron microscopic examination of the SL2 fraction. The reason forthis discrepancy remains unclear, but it should be taken into accountthat the NMR signal amplitude is proportional to the droplet volume,not to the droplet diameter. Thus, the contribution of the individuallipid droplets to the NMR spectrum is proportional to 4. This givesa strong predominance to the larger size droplets. For example, asingle droplet of5.0-@m diameter would contribute to the NMR signalto the same extent as 141 lipid droplets of 0.97-p@mdiameter.

The electron microscopy performed in this study on the C6-inducedglioma showed the presence of lipid droplets in the cytoplasm oftumoral cells in the periphery of the tumor. The presence of lipidbodies (560 nm in diameter, measured from the original figures) in theisolated C6 glioma cell line was already described by Benda et a!.(36). However, the present study shows that the number of lipiddroplets seemed to increase from the periphery toward the center ofthe tumor. No precise quantification was, however, performed. Thelarger droplets were located in damaged necrotic cells or among celldebris (Fig. 3c). The mobile lipid resonance detected in the in vivoproton spectra of the C6 glioma thus appeared to be, at least partly,due to a necrotic process. These results support those obtained byone-dimensional and two-dimensional proton spectroscopy on humanbrain tumor biopsies (12—14).Kuesel et a!. (12) demonstrated that theamount of mobile lipids could be correlated with the percentage ofnecrosis in grade IV astrocytoma samples. Furthermore, the appearance oflipid droplets or NMR-visible lipids has been described in theliterature for a variety of situations involving short-term hypoxiaand/or ischemia: acute regional heart ischemia in cats (electron microscopy evidence; Ref. 37), brain stroke in humans (‘HNMR cvidence; Refs. 38 and 39), hypoxic areas of tumors, hepatoma andEhrlich carcinoma (electron and fluorescence microscopy evidence;Ref. 40). It has been proposed that lipid droplets might represent atemporary (or end point, if tissue does not recover) storage compartment for fatty acids in the form of triglycerides. Under hypoxic orischemic conditions, fatty acids would not be degraded by (1-oxidation. Independently of the origin of such conditions, unwanted sideeffects of free fatty acid build-up in the cells would be avoided bysequestering them in the form of triglycerides in lipid droplets (for anexample, see Ref. 40). Freitas et a!. (40, 41) and Kuesel et a!. (12, 13)also suggested that mobile lipids could accumulate in viable tumorcells as a consequence of hypoxic conditions, and this would beaccelerated with the necrotic process. Recent evidence has shown theimportant role of hypoxia in the regulation of tumor development(42). On one hand, hypoxia would induce apoptosis in oncogenicallytransformed cells and thus maintain an equilibrium between cell deathand birth in the early stages of tumor growth (42). On the other hand,hypoxia selects for apoptosis-resistant cells, promoting clonal expansion of these cells, which could explain the resistance of solid tumors

to cancer therapy (42). Then, the detection of mobile lipids in tumorby ‘HMRS might be a parameter of interest to follow the development of a tumor (42), to grade the tumor (14), and for the therapeuticapproach to tumor treatment in humans (43).

In summary, the present study shows that, in the C6 experimentalrat brain tumor model, lipid droplets associated with necrosis processes constitute the major source of the mobile lipid resonancedetected in in vivo proton spectra. These results reinforce the suggestion of Kuesel et a!. (12, 13) that necrosis should be seriouslyconsidered when trying to explain the appearance of lipid patterns inhuman brain tumors in vivo.

ACKNOWLEDGMENTS

We thank Dr. A. Rosemary Tate for English language correction.

REFERENCES

I . Mountford, C. E., Grossman, G., Reid, G., and Fox, R. M. Characterization oftransformed cells and tumors by proton nuclear magnetic resonance spectroscopy.Cancer Res., 42: 2270—2276,1982

2. Mountford, C. E., and Tattersall, M. H. N. Proton magnetic resonance spectroscopyand tumor detection. Cancer Surv., 6: 285—314,1987.

3. Mountford, C. E., and Wright, L. C. Organization of lipids in the plasma membranesof malignant and stimulated cells: a new model. Trends Biochem. Sci., /3: 172—177.1988.

4. Mackinnon, W. B., Dyne, M., Holmes, K. T., Mountford, C. E., and Gupta, R. S.Further evidence that the narrow ‘Hmagnetic resonance signals from malignantcells do not arise from intracellular lipid droplets. NMR Biomed., 2: 161—164,1989.

5. Lean, C. L.. Mackinnon, w. B., and Mountford, C. E. Fucose in ‘HCOSY spectra ofplasma membrane fragments shed from human malignant colorectal cells. Magn.Reson. Med., 20: 306—311,1991.

6. Callies, R., Sri-Pathmanathan, R. M., Ferguson, D. Y. P., and Brindle, K. M. Theappearance of neutral lipid signals in the ‘HNMR spectra of a myeloma cell linecorrelates with the induced formation of cytoplasmic lipid droplets. Magn. Reson.Med., 29: 546—550,1993.

7. May, G. L., Sztelma, K., and Sorrell, T. C. The presence of cytoplasmic lipid dropletsis not sufficient to account for neutral lipid signals in the ‘HMR spectra ofneutrophils.Magn.Reson.Med..31: 212—217,1994.

8. Rémy,C., von Kienlin, M., Lotito, S., François,A., Benabid, A. L., and D&orps, M.In vivo ‘HNMR spectroscopy of an intracerebral glioma in the rat. Magn. Reson.Med., 9: 395-401, 1989.

9. Bourgeois, D., Rémy,C., La Fur, Y., Devoulon, P., Benabid, A. L., and D&orps, M.Proton spectroscopic imaging: a tool for studying intracerebral tumor models in rat.Magn. Reson. Med., 21: 10—20,1991.

10. Le Fur, Y., Ziegler, A., Bourgeois, D., Décorps,M., and Rémy,C. Phased spectrascopic images: application to the characterization of the ‘HI.3 ppm resonance inintracerebral tumors in the rat. Magn. Reson. Med., 29: 431—435, 1993.

11. Rémy,C., Anis, C., Ziegler, A., Sam Lal, E., Moreno, A., Le Fur, Y., and Décorps,M.In vivo,exvivoandinvitroone-andtwo-dimensionalnuclearmagneticresonancespectroscopy of an intracerebral glioma in rat brain: assignment of resonances.J. Neurochem.,62: 166—179,1994.

12. Kuesel, A. C., Sutherland, G. R., Halliday, W., and Smith, I. C. P. ‘HMRS of highgrade astrocytomas: mobile lipid accumulation in necrotic tissue. NMR Biomed., 7:149—155,1994.

13. Kuesel, A. C., Donnelly, S. M., Halliday, W., Sutherland, G. R., and Smith, I. C. P.Mobilelipidsandmetabolicheterogeneityof braintumoursas detectableby cxvivo‘HMRspectroscopy.NMRBiomed.,7: 172—180,1994.

14. Rutter, A., Hugenholtz, H., Saunders, J. K., and Smith, I. C. P. Classification of braintumors by cx vivo ‘HNMR spectroscopy. J. Neurochem., 64: 1655—1661,1995.

15. Posse, S., Schuknecht, B., Smith, M. E., van Ziji, P. C. M., Herschkowitz, N., andMoonen, C. T. W. Short echo time proton MR spectroscopic imaging. J. Comput.Assist. Tomogr., 17: 1—14,1993.

16. Ott, D., Hennig, J., and Ernst, T. Human brain tumors: assessment with in s'ivo protonMRspectroscopy.Radiology,186:745—752,1993.

17. Sijens, P. E., Knopp, M. V., Brunetti, A., Wicklow, K., Alfano, B., Bachert, P.,Sanders, J. A., Stillman, A. E., Kett, H., Sauter, R., and Oudkerk, M. ‘HMRspectroscopy in patients with metastatic brain tumors: a multicenter study. Magn.Reson. Med., 33: 818—826,1995.

18. Nicolay, K., van der Toom, A., and Dijkhuizen, R. M. In vivodiffusion spectroscopy:an overview. NMR Biomed., 8: 365—374,1995.

19. Benda,P.,Lightbody,J., Sato,G.,Levine,L.,andSweet,W. Differentiatedratglialcell strain in tissue culture. Science (Washington DC), 161: 370—371,1968.

20. Benabid, A. L., Rémy,C., and Chauvin, C. Experimental models of rat brain tumorsby stereotactic injection of C6 glioma and HTC hepatoma cell lines. In: M. D. Walkerand D. G. T. Thomas (eds.), Biology of Brain Tumour, pp. 221—226.Boston:Martinus Nijhoff Publishers, 1986.

21. Sun, G. Y., Huang, H. M., Kelleher, J. A., Stubbs, E. B., and Sun, A. Y. Markerenzymes, phospholipids and acyl group composition of a somal plasma membrane

413

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from

GUOMA LIPD DROPLSTS: NMR AND MICROSCOPY STUDY

fraction isolated from rat ceiebral cortex: acomparison with microsoines and synapticplasma membranes. Neurocbem. tnt., 12: 69-77, 1988.

22. Thompson, R. J. Isolation procedures and in vitro applications ofcell nuclei from themammalian brain. In: A. I. Turner and H. S. Bachelard (ads.), Neurochcmist,y: APractical Approach, pp. 225-242. Oxford: IRL Press. 1987.

23. Glauert, A. M. Fixation, dehydration and embedding ofbiological specimens. in: A.M. Glauert (ed.), Practical Methods in Electron Microscopy, Vol 3, pp. 23-31.Amsterdam: North Holland Publishing Company. 1974.

24. Williams,M. A. Quantitativemethodsin biology.In: A. M.Glauert(ad.),PracticalMethods in Electron Microscopy, Vol 6, pp. 1-84. Amsterdam: North HollandPublishing Company. 1977.

25. Weibel, E. R., and Bolender, R. P. Stereological techniques for electron microscopicmorphometry. In: M. A. Hayat (ed), Principles and Techniques of Electron Microscopy. Vol 3. pp. 237-296. New York: Van Nostrand Reinhold Company, 1974.

26. Von Kienlin,M., Décorps,M., Albrand,J. P., Foray. M. F., and Blondet,P.Hard-pulse sequence for solvent suppression which only need a zero-order phasecorrection. J. Magn. Reson., 76: 169—173,1988.

27. Décorps,M., Lebas, J. F., Leviel, J. L, Confort, S., Rémy,C., and Benabid, A. LAnalysisof brainmetabolismchangesinducedby acutepotassiumcyanideintoxication by 3'P NMR in vivo using chronically implanted surface coils. FEBS Left., 168:1—6,1984.

28. Tanner,J. E. Useof the stimulatedechoin NMRdiffusionstudies.J. Chem.Phys.,52: 2523-2526, 1970.

29. MeIbOIdt,K.D.,Hörstermann,D.,Hllnicke,W.,Bruhn,H.,andFrahm,J. Molecularself-diffusion ofintracdllularmetabolites in rat brain in viva investigated by localizedproton NMR diffusion spectroscopy. Magn. Reson. Med., 29: 125-129, 1993.

30. Sotak,C. H., and Alger,J. R. A pitfall associatedwith lactatedetectionusingstimulated-echo proton spectroscopy. Magn. Reson. Med., 17: 533—538.1991.

31. Stejskal,E. 0., andTanner,J. E. Spindiffusionmeasurements:spinechoesin thepresence of time-dependent field gradient J. Chcm. Phys.. 42: 288—292,1965.

32. Moonen, C. T. W., van Zijl, P. C. M., I.e Bihan, D., and DesPres, D. In vivo NMRdiffusion spectroscopy: 31Papplication to phosphorus metabolites in muscle. Magn.Reson. Med., 13: 467—477.1990.

33. Wick,M., Nagalomo,Y.. Priehneier,F., and Frahm,J. Alterationof intracellular

metabolite diffusion in rat brain in vivo during ischemia and reperfusion. Stroke, 26:1930—1933.1995.

34. Karasaki, S. Passage of cytoplasmic lipid into interphase nuclei in preneoplastic ratliver.J. Ultrastruct.Rca..42: 463—478,1973.

35. Cheville N. F. Cell death and degeneration. In: N. F. Cheville (ed), Cell Pathologj@,pp. 15—80.Ames: The Iowa State University Press, 1976.

36. Bends,P., Someda,K., Messer,J., and Sweet,W. H. Morphologicaland immunochemical studies of rat glial tumors and clonal strains propagated in culture. J. Neurosurg., 34: 310—323,1971.

37. Jodalen,H., Stangeland,L. Grong,K., Vik-Mo,H., and Lekven,J. Lipidaccumulation in myocardium during acute regional ischaemia in cats. J. Mol. Cell. CardioL17: 973—980.1985.

38. Saunders,D. E.. Howe,F. A., vanden Boogaart,A., Brawn,M. M.,andGriffiths,I. R. Discrimination of metabolites and lipids/macromolecules in stroke in humansusing short echo proton spectroscopy. in: Proceedings of the Third Annual Meetingof the Societyof MagneticResonance,Vol. 3, p. 1821.Nice,France:SocietyofMagnetic Resonance, 1995.

39. Hwang,J. H., Graham,G. D., Behar,K. L, Alger,J. R., Prichard,J. W., andRothman, D. L Short echo time proton magnetic resonance spectroscopic imaging ofmacromolecule and metabolite signal intensities in the human brain. Magn. Reson.Med.,35: 633—639,1996.

40. Freitas, I., Pontiggia, P., Barni, S., Bertone, V., Parente, M., Novarina, A., Roveta, G.,Gerzei, G., and Stoward, P. Histochemical probes for the detection of hypoxictumour cells. Anticancer Res., 10: 613-622, 1990.

41. Freitas,I.,Griffith,P.,Bono,B.,Barni,S.,Bertone,V.,Baronzio,G.F.,andGerzeli,0. Characterization ofthe metabolism ofperinecrotic cells in solid tumors by enzymehistochemistry. in: Abstracts of the Fifth InternatiOnal Conference of AnticancerResearch, pp. 1703-1704. Corfu, Greece: International Institute of Anticancer Research, 1995.

42. Graeber,T. G.,Osmanian,C.,Jacks,T., Housman,D. E.,Koch,C. J., Lowe,S. W.,and Giaccia, A. J. Hypoxia-mediated selection of cells with diminished apoptolicpotential in solid tumours. Nature (Lond.), 379: 88-91, 1996.

43. Hill,R.P. Cellularbasisof radiotherapy.in: 1.F.TannockandR.P. Hill(eds.).TheBasicScienceof Oncology,pp. 237-255.NewYork:PergamonPress,1987.

414

American Association for Cancer Research Copyright © 1997 on July 13, 2011cancerres.aacrjournals.orgDownloaded from