Journal of Neuro-Oncology 62: 329–338, 2003.© 2003 Kluwer Academic Publishers. Printed in the Netherlands.

Clinical Study

Evaluation of brain tumor metabolism with [11C]choline PET and 1H-MRS

M. Utriainen1,2, M. Komu3, V. Vuorinen4, P. Lehikoinen1, P. Sonninen3, T. Kurki3, T. Utriainen7, A. Roivainen1,H. Kalimo5 and H. Minn1,6

1Turku PET Centre; 2Department of Pediatrics; 3Department of Radiology; 4Department of Neurosurgery;5Department of Pathology; 6Department of Oncology and Radiotherapy, University of Turku, Turku;7Department of Oncology, University of Helsinki, Helsinki, Finland

Key words: brain tumor, tumor metabolism, nuclear magnetic resonance imaging, positron emission tomography,[11C]choline

Summary

Background: The signal of choline containing compounds (Cho) in proton magnetic resonance spectroscopy(1H-MRS) is elevated in brain tumors. [11C]choline uptake as assessed using positron emission tomography (PET)has also been suggested to be higher in brain tumors than in the normal brain. We examined whether quantitativeanalysis of choline accumulation and content using these two novel techniques would be helpful in non-invasive,preoperative evaluation of suspected brain tumors and tumor malignancy grade.

Methods: 12 patients with suspected brain tumor were studied using [11C]choline PET, gadolinium enhanced 3-Dmagnetic resonance imaging and 1H-MRS prior to diagnostic biopsy or resection. Eleven normal subjects served ascontrol subjects for 1H-MRS.

Results: The concentrations of Cho and myoinositol (mI) were higher and the concentration of N-acetylsignal/group (NA) lower in brain tumors than in the corresponding regions of the normal brain. There were no sig-nificant differences in metabolite concentrations between low- and high-grade gliomas. In non-tumorous lesionsCho concentrations were lower and NA concentrations higher than in any of the gliomas. Enormously increasedlipid peak differentiated lymphomas from all other lesions. The uptake of [11C]choline at PET did not differ betweenlow- and high-grade gliomas. The association between Cho concentration determined in 1H-MRS and [11C]cholineuptake measured with PET was not significant.

Conclusion: Both 1H-MRS and [11C]choline PET can be used to estimate proliferative activity of human braintumors. These methods seem to be helpful in differential diagnosis between lymphomas, non-tumorous lesions andgliomas but are not superior to histopathological methods in estimation of tumor malignancy grade.

Introduction

Carcinogenesis and cancer growth are characterized byenhanced cell proliferation and, therefore, by enhancedcell membrane synthesis. Lecithin, a major phospho-lipid constituent of the cell membrane, is synthesizedfrom choline. Several studies using proton magneticresonance spectroscopy (1H-MRS) have shown ele-vated signal of choline containing compounds (Cho)(i.e. phosphorylcholine and glycerophosphocholine) inbrain tumors [1–3].

Circulating choline is the major source of choline inthe brain. While normal brain cells are at non-dividing

stage, the uptake of choline by normal brain hasbeen supposed to be low [4]. Indeed, the accumula-tion of [11C]-labeled choline in normal brain tissue isextremely low both in rodents [5] and humans [6].This is in contrast to several non-neuronal tissues suchas kidney, liver, lung and pancreas [5].

Recently, an interest to assess [11C]choline uptakein human tumors with positron emission tomography(PET) has been raised. Hara et al. [6] studied patientswith prostate cancer and found that in normal and can-cerous prostate tissue accumulation of [11C]choline waseasier to detect than that of 18F-fluoro-deoxyglucose([18F]FDG) because of low background activity of

330

[11C]choline in bladder and other surrounding pelvictissues. The same authors also studied patients withbrain tumors and concluded that clear delineation ofprimary and/or residual tumors from the surroundingbrain tissue is the major advantage of [11C]choline PET[7,8]. High tumor-to-background radioactivity ratiowas found to characterize all primary and metastaticbrain tumors studied while non-tumorous lesions, i.e.infarct or hemorrhage, did not accumulate [11C]choline.However, their analysis did not differentiate malig-nant and benign tumors [7] whereas PET using both[18F]FDG and [11C]methionine ([11C]MET) has beenshown to make such differentiation possible [9–11].The distribution of tumor grade and histologic subtypeswas also slightly biased in the study by Hara et al. [7,8]since no low-grade gliomas i.e. grade II astrocytomaswere included despite their common occurrence.

1H-MRS enables quantitation of several tissue con-stituents such as Cho, N-acetyl signal/group (NA),creatinine (Cr) and myoinositol (mI) non-invasively inthe brain. It has been shown that spectra of tumorsrevealed by 1H-MRS differ from those of the normalbrain. Typical changes in spectra include an increase inCho, a decrease in NA, and detection of lactate or lipidpeaks in high-grade tumors [12].

The present study was undertaken to examine theaccumulation of [11C]choline and concentration of Choin suspected brain tumors using PET and 1H-MRS,respectively. Patient selection was based on clini-cal findings including contrast-enhanced computedtomography (CT) and magnetic resonance imaging(MRI) and the diagnosis was subsequently verified his-tologically in each case. Our goal was also to studywhether quantitative analysis of choline accumulation

and content would be helpful in non-invasive eval-uation of tumor malignancy grade. Specifically, wewanted to compare two methods, [11C]choline PETand 1H-MRS, and their sensitivity to characterizeneoplastic differentiation.

Patients and method

Patients

We studied 12 patients (median age 53 years, range26–72 years) with a suspicion of brain tumor. Charac-teristics of the patients are given in Table 1. To alleviateintracranial pressure, corticosteroid therapy had beenstarted in all patients. In addition to these 12 patients,11 control subjects were studied using 1H-MRS.

Study design

Each patient underwent [11C]choline PET, gadoliniumenhanced 3-D MRI and 1H-MRS prior to any oncologictreatment. All the three procedures were performedwithin two days. After PET and MRI, each patient hadsurgery for histopathological verification of diagno-sis. WHO classification [13] was used for grading oftumors and immunohistochemical methods were thenapplied as explained in the section ‘Histochemistry’.

Ethics

The study was reviewed and approved by the JointCommittee on Ethics of the Turku University and theTurku University Central Hospital. The nature, purpose

Table 1. Characteristics of the patients

Sex Age PAD Ki-67MIB-1 Visual accumulation Ki (×10−2) SUV Choline(years) (%) of [11C]choline (min−1) concentration

M 53 Non-tumor lesion 0 − 0.4 0.25 2.69F 69 Demyelinizating disease 0 − 2.7 0.62 2.76M 26 Astrocytoma gr II 5.7 − 0.5 0.07M 31 Oligoastrocytoma gr II 10.3 − 0.5 0.22 5.00M 34 Astrocytoma gr II 10.0 − 0.6 0.09 2.87F 35 Astrocytoma gr II 2.6 + 1.3 0.99 3.30F 66 Astrocytoma gr II 16.2 + 7.8 3.31 5.21F 46 Astrocytoma gr III 3.1 + 0.2 0.17 3.76M 57 Astrocytoma gr III 18.5 + 2.3 1.13 3.80M 72 Glioblastoma multiforme 9.1 + 4.5 2.34 3.73M 68 Lymphoma metastasis 45.1 + 7.7 3.50M 69 Large cell lymphoma metastasis 36.8 + 2.2 0.79

331

and potential risks of the study were explained to allsubjects before they gave their informed consent toparticipate.

PET procedure

Patient preparation and scanner performanceThe PET studies were performed by using an 18-ringGE advance whole body tomograph (General ElectricMedical Systems, Milwaukee, WI, USA). GE advancehas an axial and spatial resolution of 4.3 mm.

Tracer synthesisThe radioactive [methyl-11C]choline was synthesizedfrom dimethylaminoethanol by direct methylationwith [11C]methyl triflate at the RadiopharmaceuticalChemistry Laboratory of the Turku PET Centre asdescribed earlier [5]. Labeled choline was then puri-fied by means of solid phase extraction, and finallydissolved in saline. The radiochemical purity exceeded99.9% and the average specific radioactivity was80 GBq/µmol at the end of synthesis.

Blood sampling, image acquisition and processingTwo catheters were inserted; one in an antecubital veinfor infusion of [11C]choline, and one in the oppositeradial or brachial artery for blood sampling. Beforeemission scan, a 8 min transmission scan for correctionof photon attenuation was performed with a pair of pinsources containing 68Ge/68Ga. After a bolus injectionof [11C]choline, a dynamic emission scan was acquiredfor 40 min. The entire brain could be scanned with a15 cm axial field of view of the tomograph.

Image processingAll data were corrected for decay and measured photonattenuation and reconstructed into 128 × 128 matrix.The final in-plane resolution in reconstructed andHann-filtered images was 7.4 mm.

Data analysisVisual analysis of PET images was performed bythree independent observers. For quantitative analysisof [11C]choline uptake, the graphical analysis methoddescribed by Patlak et al. [14] was used. This methodwas originally developed for [18F]FDG uptake buthas been generally applied for other tracers such as[11C]MET [15,16]. Accumulation of the tracer in tissueat a fixed time-point was also measured as standardizeduptake value (SUV), i.e. the radioactivity concentration

in the tumor regions of interest (ROI) per injected[11C]choline dose corrected by the body weight.

Manually contoured ROIs were defined on the wholetumor area by visual comparison with the correspond-ing contrast-enhanced MR image on PET imagessummed over frames corresponding time intervalbetween 20 and 40 min after injection of [11C]choline.A semi-automatic software was applied for defining thehighest tracer accumulation (3 × 3 pixels) within thewhole tumor ROI (ROImax).

MRI and MRS

A commercial 1.5 T MR imager (GE Signa HorizonLX with the software version of 8.2, General Elec-tric Medical Systems, Milwaukee, WI, USA) with aquadrature transmitting/receiving bird cage head coilwas used for 1H-MRS investigations. Transverse fastspin echo T2-weighted (T2W) MR images (TR/TE =4500/85 ms, ETL = 12, slice thickness = 5 mm,FOV = 260 mm, matrix size = 224 × 512) werefirst acquired. The single voxel was placed on theimages using the transversal T2W-images and the addi-tional coronal and transversal single shot fast spin echoT2W-images.

An automatic signal acquisition and processingpackage PROBE/SV [17] was used to acquire avolume-localized, water-suppressed spectrum from asingle voxel in the patients. The nuclear magnetic res-onance (NMR) signal was collected using a PRESSsequence PROBE-P, including the automated adjust-ment of field homogeneity through the voxel andwater suppression with the parameters TR/TE =6000/35 ms, number of excitations 8, total numberof scans 32, and acquisition time of 5 min 48 s. Thevolume of the voxel varied from 5.0 to 8.9 ml. Withthe Auto Prescan period, the total scanning time of aspectrum was about 10 min. To recognize the possi-ble lactate signal, the PRESS sequence was repeated,if necessary, with the same parameters with the excep-tion of TR/TE = 2000/144 ms. The default value ofthe body temperature was 37◦C. Immediately after thepatient or control subject study, the NMR signal fromthe GE MRS phantom was collected with the sameparameters as in the first PRESS sequence with theexception of the temperature. The temperature sensi-tive strip indicated the temperature of the phantom. Thetemperature of the MRS phantom was 22 ± 0.5◦C.

The PROBE/SV reconstruction produces the outputinvolving an image containing the real part fast Fourier

332

transformation spectrum and quantification values andratios for four metabolites: NA, Cho, Cr and mI, unsup-pressed water and noise [18]. The basic metabolicpeaks for NA, Cho and Cr are defined at the chem-ical shift of 2.02, 3.03 and 3.22 ppm, respectively.The quantification of the metabolic peak succeeds if thesignal-noise-ratio of the peak is good enough and thepeak is found in the line region. The effective areas ofthe NA, Cho, Cr, mI and water peaks were recordedfor the brain and for the reference phantom. The con-centrations of the metabolites NA, Cho, Cr and mIwere calculated after several corrections using the GEMRS phantom simulating the human head [19]. Thepatients and control subjects were divided into twogroups according to their age. The first group consistedof patients and control subjects of 30–60 years and thesecond group of those 60–80 years. The mean value ofT2-saturation was calculated separately for the controlgroups I and II and these values were used to correctthe T2-saturation of the corresponding patient group.

By using the corrected values of the peak areas of thebrain and the phantom and the known phantom con-centrations, the absolute concentrations of NA, Cho,Cr and mI within the selected voxel of brain werecalculated for patients and control subjects.

Histochemistry

Proliferation activityKi-67 labeling was performed by using monoclonalanti-Ki-67 (MIB-1, IgG1 subclass, Immunotech S.A.,Marceille, France) and an avidin–biotin immunoperox-idase technique as described previously [20,21], withslight modifications. Briefly, the five-µm-thick paraf-fin embedded sections of tumors were incubated in cit-rate buffer during microwave oven heating. The boundantibody was visualized by using diaminobenzidine aschromogen.

The Ki-67MIB-1 indices were quantified visually bypoint-counting on tumor areas expressing subjectivelythe highest number of immunopositive nuclei. Theassessment proceeded in adjacent microscopic fieldsof a ×25 objective lens along horizontal and verti-cal axes perpendicular to each other until 1000 cellswere counted. Only neoplastic cells were included inthe quantification of Ki-67-positive cells. Necrotic andhemorrhagic areas and the section borders were alsoomitted from quantification. The results are expressedas the percentage of Ki-67-positive cells per 1000tumor cells.

Cell densityThe assessment of cell density was performed in thesections immunostained to visualize Ki-67MIB-1. Neo-plastic and nonneoplastic cells were quantified usinga ×25 objective lens along a horizontal and a verti-cal axis perpendicular to each other until 1000 cellswere counted. Endothelial cells were not analyzed. Theresults are expressed as the number of cells per mm2.

Statistical methods

A simple factorial ANOVA was used to test the signif-icance of the difference between the concentrations ofNA, Cho, Cr and mI of the patients and control subjects.Due to the skewed distributions other group compar-isons were performed using the Mann–Whitney test.Spearman’s correlation coefficients were calculatedto assess the possible association between metabo-lite concentrations in tumors measured with 1H-MRS,11C-choline uptake rate and histochemical data. All sta-tistical analyses were performed using the SPSS soft-ware (SPSS Inc. Chicago, IL). The results are expressedas mean±SD. Two-tailed p-values less than 0.05 wereconsidered statistically significant.

Results

1H-magnetic resonance spectroscopy

Cho concentration was significantly higher in braintumors than in the corresponding areas of the brain ofthe normal subjects (3.68 ± 0.92 vs. 2.31 ± 0.44 mM,p = 0.003). Likewise, the concentration of mI wassignificantly higher in patients as compared to normalsubjects (12.44±2.70 vs. 7.98±2.73 mM, p = 0.023).The concentration of NA was significantly smallerin patients than in normal subjects (3.93 ± 3.12 vs.10.75 ± 2.11 mM, p < 0.001) whereas there was nodifference between the concentration of Cr (8.62±1.76vs. 8.94 ± 1.72 mM, NS).

Two representative spectra and the correspondingMR and [11C]choline images of patients with low-gradeand high-grade gliomas are shown in Figures 1 and 2,respectively. The spectral patterns of gliomas weresimilar to those previously published, and Cho concen-tration averaged 3.95±0.32 mM. There were no signif-icant differences in the concentrations of NA, Cho, Cror mI between grade II and grade III gliomas. Specif-ically, Cho concentrations averaged 4.09±0.59 mM in

333

(a)

(b)

(c)

the low-grade and 3.77 ± 0.02 mM in the high-gradegliomas (NS). On the other hand, the two patients withlymphoma (a primary CNS lymphoma and a metastaticlymphoma) presented with very different spectra incomparison to those found in the patients with gliomas.A representative spectrum of CNS lymphoma is shownin Figure 3. In lymphomas the lipid peak was enor-mously increased, which even prevented the quantita-tion of all the other peaks, i.e. NA, Cho, Cr and mI. Choconcentrations in the two non-tumorous lesions (2.69and 2.75 mM) were lower than in any of the gliomas(range 2.87–5.21 mM) while the concentration of NAwas higher in non-tumorous lesions than in gliomas(8.55 ± 0.99 vs. 2.60 ± 0.70 mM, p < 0.01).

[11C]choline PET

Visual analysis of [11C]choline accumulationIn the visual analysis of [11C]choline PET studies7 out of 12 tumors were graded as hypermetabolic.[11C]choline accumulation was clearly higher than thatof the surrounding brain tissue in all high-grade tumors(n = 5). In addition, two out of five low-grade tumorsaccumulated [11C]choline whereas neither of the twonon-neoplastic lesion showed any [11C]choline accu-mulation. There was no disagreement between theinterpretations of the three observers in any case.

Graphically obtained Ki and SUV determined inthe steady state gave highly comparable results. Ki

was positively associated with both SUV (r = 0.98,p < 0.001) and SUVmax (r = 0.96, p < 0.001) in thewhole population. Patlak Ki for net [11C]choline uptakein gliomas averaged 0.022 ± 0.009 min−1, and did notdiffer between low- and high-grade gliomas (0.021 ±0.014 vs. 0.022±0.012 min−1, NS). [11C]choline SUVswere also similar in both low- and high-grade groups(0.94 ± 0.62 vs. 1.21 ± 0.63, NS). Both Patlak Ki andSUVs tended to be higher in the two CNS lymphomasthan in the gliomas (Table 1).

The association between Cho concentration deter-mined in 1H-MRS and [11C]choline uptake mea-sured with PET did not achieve statistical significance

Figure 1. The representative MRI (a), spectrum (b) and[11C]choline PET image (c) of a patient with low-grade gliomalocated frontoparietally on the right-hand side. The followingmetabolites were detected: mI at 3.56 ppm, Cho at 3.21 ppm, Cr at3.04 ppm and N-acetyl aspartate (NA) at 2.02 ppm in the protonspectrum. In the spectrum, the choline peak is increased. Therewas no clear extra choline accumulation in the PET scan.

334

(a)

(b)

(c)

irrespective of whether Patlak Ki or SUVs were usedin analysis (r = 0.50–0.54, p = 0.1).

Proliferative activity

Ki-67MIB index tended to be higher in the high-gradethan in the low-grade gliomas (10.2±4.5 vs. 8.9±2.3,NS). In the two lymphomas the proliferative index wasmuch higher (36.8 and 45.1) than in any of the gliomas(range 2.6–18.5) while, as expected, no proliferativeactivity was found in the non-tumorous lesions.

Ki-67MIB was positively associated with both Choconcentration as measured with 1H-MRS and with Choaccumulation as measured with PET (r = 0.59–0.65,p = 0.04–0.05).

Discussion

The present study was undertaken to assess the accumu-lation of [11C]choline and concentration of Cho in sus-pected malignant brain tumors using PET and 1H-MRS,respectively. We applied a dynamic method originallydeveloped for 18FDG to quantitate [11C]choline netinflux in tumor tissue and compared this approach toa more simple measurement of regional tracer con-centration (SUV) to study whether quantitation of[11C]choline uptake would be useful in differentia-tion of various tumors and tumor malignancy gradeas has been suggested by previous investigators [6,7].Finally, we assessed with two independent methodswhether the concentration of Cho bears a relationshipto the rate of choline uptake in brain tumors. Unfor-tunately, we did not find evidence that low- and high-grade gliomas could be differentiated based on theircholine metabolism by using either of the two novelnon-invasive methods. This was true even when bloodmetabolism of choline was taken into account in theanalysis of [11C]choline PET studies. By contrast, lym-phomas could be clearly characterized by both highcholine concentrations and [11C]choline uptake. Non-tumorous brain lesions, on the other hand, could bedifferentiated from gliomas based on the lower contentof choline and higher content of NA as compared withtumors. While interpreting these data one must keep in

Figure 2. The representative MRI (a), spectrum (b) and[11C]choline PET image (c) of a patient with high-grade gliomain right frontal lobe. There is an intensive choline accumulationin the tumor area in the PET image.

335

(a)

(b)

(c)

mind that our study population was rather small due tothe heavy scanning protocols in severely ill patients.

Our study material consists of 12 consecutivepatients having a suspected brain tumor which neces-sitated a diagnostic biopsy. The patients were enrolledfor this study in a single institution based on their clin-ical findings. Two patients were ultimately found tohave non-neoplastic lesions. In our view, this diagnosticuncertainty further emphasizes the need to develop newnon-invasive tools for preoperative characterization ofbrain tumors.

The present study confirms some key findings ofprevious studies on human brain tumors. The concen-tration of Cho is increased in brain tumors [3,22,23]but choline concentration does not necessarily seemto be associated with the malignancy grade [12]. Con-centration of NA which is thought to reflect neuronaldensity and viability has also been found to be lowerin brain tumors suggesting neuronal loss. Interest-ingly, our population also included two patients withCNS lymphomas which have seldom been studied with1H-MRS previously [24,25]. These tumors had highlycharacteristic spectra which could easily be differen-tiated from normal brain tissue, non-tumorous lesionsand gliomas. The high lipid peak in lymphomas mightreflect the high amount of necrotic cells in these tumors.Finally, it seems that non-tumorous lesions can fairlyreliably be differentiated from gliomas and lymphomasbased on their characteristic spectra showing higher NAand lower choline concentrations than their neoplasticcounterparts.

We have recently examined the kinetic behav-ior of [11C]choline in rats and humans [5]. In thepresent study, arterial rather than venous plasma sam-ples were used for the measurement of [11C]cholineradioactivity concentration. No model for quantitationof [11C]choline uptake in tumor tissue has thus farbeen published. In previous studies, Hara et al. haveexpressed [11C]choline accumulation using tumor-to-reference area radioactivity concentration [6–8].This method might not be ideal owing to exten-sive metabolism of [11C]choline in normal brain, andpossibly wide interindividual variation in metaboliccapacity. Therefore, we employed the graphical anal-ysis method of Patlak et al. [14] originally developed

Figure 3. The representative MRI (a), spectrum (b) and[11C]choline PET image (c) of a patient with CNS lymphomalocated in right temporal lobe. In the spectrum the lipid peak isenormously increased.

336

for [18F]FDG uptake but successfully adopted for[11C]MET [15,16] and found it to be applicable also forquantitation of [11C]choline uptake. However, a reliablemeasurement of [11C]-labeled metabolites is a prereq-uisite for this method as recently discussed in detail [5].

While we envisaged that the use of kinetic ratherthan steady-state method would prove to be more use-ful, we were surprised to find that both methods gavehighly comparable results. This suggests that arterialblood sampling, quantitation of Cho metabolites andgraphical analysis of the results do not necessarily out-weigh more simple and less invasive and laboriousmethods such as SUVs at least in the clinical setting.However, the SUV-based approach we employed isnot identical to that used by Hara and coworkers [7,8]since SUV only account for radioactivity concentrationwithin the region of interest corrected for body weightand injected dose while Hara’s approach included ref-erence areas in seemingly normal brain tissue. Due tothe very low accumulation of [11C]choline in the normalbrain the anatomical resolution of normal brain struc-tures is poor and definition of reference areas may notalways be simple. We would therefore recommend theuse of SUV-based approach for further studies at leastin clinical studies in brain tumor patients. However,larger studies are necessary for direct comparison ofthe different quantitative approaches.

[11C]choline uptake did not differentiate betweenlow- and high-grade gliomas. This is in contra-diction with a recent study by Ohtani et al. [26]who studied brain tumors using MRI, [11C]choline-and [18F]FDG-PET. They found higher tumor/whitematter count ratios for [11C]choline in high-gradegliomas compared to low-grade gliomas in a sampleof 14 glioma patients. Surprisingly they had, however,excluded a patient with intensively accumulating pilo-cytic astrocytoma from the analyses. As already dis-cussed, the use of tumor/normal brain count ratios inanalyses might not be ideal. Based on the present studyand previous ones from a single institution [6,7] thevariation in [11C]choline uptake seems to be highlyvariable even among tumors with same histologicaldiagnosis. Most if not all high-grade tumors showvisually high [11C]choline accumulation but some low-grade tumors such as grade II astrocytomas, pituitaryadenomas and meningeomas may also exhibit compa-rable [11C]choline accumulation. Interindividual vari-ation of choline uptake in the adjacent normal brainmight also explain the discrepancy between visuallyobserved [11C]choline accumulation and quantitated

[11C]choline uptake. Although [11C]choline PET maynot be suitable for estimation of tumor grade it may beuseful for delineation of tumor extent. This might behelpful e.g. in radiotherapy planning. There are, how-ever, other tracers such as [11C]MET which seem to beequally suitable for delineation of tumor extent [27].In addition, [11C]MET PET may be helpful in tumorgrading [11,28,29] and simultaneously provides betterresolution from adjacent brain structures as comparedto [11C]choline.

Ki-67MIB is an index commonly used to character-ize proliferative activity of variety of tumors includ-ing cerebral gliomas [30–32]. Ki-67MIB has been shownto be associated e.g. with grade of malignancy andprognosis. The uptake and intracellular concentrationof choline, a precursor of a major phospholipid con-stituent of cell membranes, lecithin, might be expectedto be increased in rapidly proliferating tumors. Infact, both the concentration of Cho and [11C]cholineuptake were positively and significantly associatedwith Ki-67MIB. Both choline concentration and uptaketherefore provide some information on tumor prolifer-ative capacity although this may not be sufficient fortumor grading. Finally, it remains to be shown whichof the studied tumor features i.e. histological grade,Ki-67MIB index, concentration of Cho or [11C]cholineuptake has the highest prognostic significance in termsof survival.

Association between the concentration of Cho mea-sured with 1H-MRS and [11C]choline uptake measuredwith PET was not significant. It is uncertain whethersuch an association should be expected since the con-centration of Cho represents intracellular metabolitepools of phosphocholine and glycerophosphocholinewhereas the rate of [11C]choline uptake is thought tobe controlled by amino acid transporter expression anddensity in tumor endothelial cells [33]. One could alsospeculate that highly proliferative cancer cells are likelyto rapidly incorporate choline and its derivatives such aslecithin to cell membranes, which explains the fact thatthe concentration of Cho is not necessarily associatedwith increased [11C]choline uptake.

In PET studies, the region of interest can be freelydefined by the investigator in emission images. Toobtain a high-quality spectra in 1H-MRS, we wereoccasionally forced to use a reasonable large voxel,i.e. 3-D volume in which the measurement was done.In some cases also normal and/or necrotic tissue inrapidly growing tumors may have been included withinthe voxel. Therefore, the results do not necessarily

337

represent the characteristics of the most rapidly pro-liferating part of the tumor. The same drawback alsoapplies to surgical or stereotactic biopsy since the his-tological sample may not perfectly represent the mostmalignant cell population of the whole tumor. Forexample, necrotic tissue may be sampled in a veryrapidly proliferating tumor or a sample consisting ofseemingly benign cells may be obtained in a heteroge-nous tumor.

In conclusion, both 1H-MRS and [11C]choline PETcan be used to estimate proliferative activity ofhuman brain tumors. However, these methods arenot superior to conventional neuroimaging methods.Specifically, marked overlap between various tumorsand tumor types does not allow estimation of tumormalignancy grade using these methods. However, lym-phomas and non-tumorous lesions could be differenti-ated from gliomas, and some diagnostic biopsies mightbe avoided if 1H-MRS was more generally available.

Acknowledgements

This work was supported by the Cancer Society ofFinland and the Ida Montin Foundation.

References

1. Bruhn H, Frahm J, Gyngell ML, Merboldt KD, Hanicke W,Sauter R, Hamburger C: Non-invasive differentiation oftumors with use of localized H-1 MR spectroscopyin vivo: initial experience in patients with cerebral tumors.Radiology 172: 541–548, 1989

2. Usenius JP, Kauppinen RA, Vainio PA, Hernesniemi JA,Vapalahti MP, Paljarvi LA, Soimakallio S: Quantitativemetabolite patterns of human brain tumors: detection by 1HNMR spectroscopy in vivo and in vitro. J Comput AssistTomogr 18: 705–713, 1994

3. Preul MC, Caramanos Z, Collins DL, Villemure JG,Leblanc R, Olivier A, Pokrupa R, Arnold DL: Accurate,noninvasive diagnosis of human brain tumors by using pro-ton magnetic resonance spectroscopy. Nat Med 2: 323–325,1996

4. Friedland RP, Mathis CA, Budinger TF, Moyer BR,Rosen M: Labeled choline and phosphorylcholine:body distribution and brain autoradiography: concisecommunication. J Nucl Med 24: 812–815, 1983

5. Roivainen A, Lehikoinen P, Gronroos T, Forsback S,Kahkonen M, Minn H: Blood metabolism of [methyl-11C]choline; implications for in vivo imaging with positronemission tomography. Eur J Nucl Med 27: 25–32, 2000

6. Hara T, Kosaka N, Kishi H: PET imaging of prostate can-cer using carbon-11-choline. J Nucl Med 39: 990–995,1998

7. Shinoura N, Nishijima M, Hara T, Haisa T, Yamamoto H,Fujii K, Mitsui I, Kosaka N, Kondo T: Brain tumors: detec-tion with C-11 choline PET. Radiology 202: 497–503,1997

8. Hara T, Kosaka N, Shinoura N, Kondo T: PET imagingof brain tumor with [methyl-11C]choline. J Nucl Med 38:842–847, 1997

9. Di Chiro G: Positron emission tomography using [18F] fluo-rodeoxyglucose in brain tumors. A powerful diagnostic andprognostic tool. Invest Radiol 22: 360–371, 1986

10. Delbeke D, Meyerowitz C, Lapidus RL, Maciunas RJ,Jennings MT, Moots PL, Kessler RM: Optimal cutoff lev-els of F-18 fluorodeoxyglucose uptake in the differentia-tion of low-grade from high-grade brain tumors with PET.Radiology 195: 47–52, 1995

11. Herholz K, Holzer T, Bauer B, Schroder R, Voges J,Ernestus RI, Mendoza G, Weber-Luxenburger G, Lottgen J,Thiel A, Wienhard K, Heiss WD: 11C-methionine PET fordifferential diagnosis of low-grade gliomas. Neurology 50:1316–1322, 1998

12. Fulham MJ, Bizzi A, Dietz MJ, Shih HH, Raman R,Sobering GS, Frank JA, Dwyer AJ, Alger JR, Di Chiro G:Mapping of brain tumor metabolites with proton MRspectroscopic imaging: clinical relevance. Radiology 185:675–686, 1992

13. Kleihues P, Cavenee WK: Pathology & Genetics ofTumours of the Nervous System. World Health Organiza-tion Classification of Tumours. IARC Press, Lyon, France,2000

14. Patlak CS, Blasberg RG: Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data.Generalizations. J Cereb Blood Flow Metab 5: 584–590,1985

15. Hatazawa J, Ishiwata K, Masatoshi I, Kameyama M,Kubota K, Ido T, Matsuzawa T, Yoshimoto T, Watanuki S,Seo S: Quantitative evaluation of L-[Methyl-C-11] methio-nine uptake in tumor using positron emission tomography.J Nucl Med 30: 1809–1813, 1989

16. Ericson K, Blomqvist G, Bergstrom M, Eriksson L,Stone-Elander S: Application of a kinetic model on themethionine accumulation in intracranial tumours studiedwith positron emission tomography. Acta Radiologica 28:505–509, 1987

17. Webb PG, Sailasuta N, Kohler SJ, Raidy T, Moats RA,Hurd RE: Automated single-voxel proton MRS: technicaldevelopment and multisite verification. Mag Reson Med 31:365–373, 1994

18. Probe Reconstruction and Quantitation Calculations. GEMedical Systems, Milwaukee, Wisconsin, 1998

19. Danielsen ER, Henriksen O: Absolute quantitative pro-ton NMR spectroscopy based on the amplitude of thelocal suppression pulse. Quantification of brain water andmetabolites. NMR Biomed 7: 311–318, 1994

20. Sallinen PK, Haapasalo HK, Visakorpi T, Helen PT,Rantala IS, Isola JJ, Helin HJ: Prognostication of astrocy-toma patient survival by Ki-67 (MIB-1), PCNA, and S-phase

338

fraction using archival paraffin-embedded samples. J Pathol174: 275–282, 1994

21. Haapasalo HK, Sallinen PK, Helen PT, Rantala IS, Helin HJ,Isola JJ: Comparison of three quantitation methods forPCNA immunostaining: applicability and relation to sur-vival in 83 astrocytic neoplasms. J Pathol 171: 207–214,1993

22. Tedeschi G, Lundbom N, Raman R, Bonavita S, Duyn JH,Alger JR, Di Chiro G: Increased choline signal coincidingwith malignant degeneration of cerebral gliomas: a serialproton magnetic resonance spectroscopy imaging study.J Neurosurg 87: 516–524, 1997

23. Hwang JH, Egnaczyk GF, Ballard E, Dunn RS, Holland SK,Ball WS, Jr.: Proton MR spectroscopic characteristics ofpediatric pilocytic astrocytomas. AJNR Am J Neuroradiol19: 535–540, 1998

24. Kinoshita Y, Yokota A: Absolute concentrations of metabo-lites in human brain tumors using in vitro proton magneticresonance spectroscopy. NMR Biomed 10: 2–12, 1997

25. Kuhlmann T, Schroter A, Dechent P, Weber F, Rustenbeck H,Fuzesi L, Bruck W, Ehrenreich H, Frahm J: Diagnosis of amultifocal B cell lymphoma with preceding demyelinatingcentral nervous system lesions by single voxel proton MRspectroscopy. J Neurol Neurosurg Psychiatry 70: 259–262,2001

26. Ohtani T, Kurihara H, Ishiuchi S, Saito N, Oriuchi N,Inoue T, Sasaki T: Brain tumor imaging with carbon-11choline: comparison with FDG PET and gadolium-enhanced MR imaging. Eur J Nucl Med 28: 1664–1670,2001

27. Utriainen M, Metsahonkala L, Salmi T, Utriainen T,Kalimo H, Pihko H: Metabolic characterization of child-hood brain tumors: comparison of 18F-fluorodeoxyglucose

and 11C-methionine positron emission tomography. Cancer95: 1376–1386, 2002

28. Ogawa T, Shishido F, Kanno I, Inugami A, Fujita H,Murakami M, Shimosegawa E, Ito H, Hatazawa J,Okudera T, Uemura K, Yasui N, Mineura K: Cerebralglioma: evaluation with methionine PET. Radiology 186:45–53, 1993

29. Nuutinen J, Leskinen S, Lindholm P, Soderstrom KO,Någren K, Huhtala S, Minn H: Use of carbon-11 methioninepositron emission tomography to assess malignancy gradeand predict survival in patients with lymphomas. Eur J NuclMed 25: 729–735, 1998

30. Haapasalo H, Sallinen S, Sallinen P, Helen P, Jaaskelainen J,Salmi TT, Paetau A, Paljarvi L, Visakorpi T, Kalimo H: Clin-icopathological correlation of cell proliferation, apoptosisand p53 in cerebellar pilocytic astrocytomas. NeuropatholAppl Neurobiol 25: 134–142, 1999

31. Rhodes RH: Biological evaluation of biopsies from adultcerebral astrocytomas: cell-growth/cell-suicide ratios andtheir relationship to patient survival. J Neuropathol ExpNeurol 57: 746–757, 1998

32. Ho DM, Wong TT, Hsu CY, Ting LT, Chiang H: MIB-1labeling index in nonpilocytic astrocytoma of childhood: astudy of 101 cases. Cancer 82: 2459–2466, 1998

33. Miyagawa T, Oku T, Uehara H, Desai R, Beattie B,Tjuvajev J, Blasberg R: Facilitated amino acid transport isupregulated in brain tumors. J Cereb Blood Flow Metab 18:500–509, 1998

Address for offprints: Meri Utriainen, Department of Oncology,P.O. Box 180, FIN-00029 HUCH, Helsinki, Finland; Tel.: +358-9-4711; Fax: +358-9-47174201; E-mail: meri.utriainen@hus.fi