Introduction

Positron emission tomography (PET) is a well-established modality in the evaluation of brain tumour patients. Twenty-two years ago the first brain tumour images were acquired with 18F-fluoro-2-deoxyglucose (FDG), which has become the most widely used tracer in PET imaging [1]. Although the usefulness of FDG in tumour imaging is beyond doubt, there are drawbacks. The high physiological uptake in the brain often renders it difficult to differentiate tumour from normal brain tissue, a problem which cannot always be overcome by correlation with structural imaging. Another problem is the large range of FDG uptake in different brain tumours, leading to an overlap with the degree of uptake in benign lesions [2, 3]. Furthermore, FDG is taken up by tumour cells as well as inflammatory cells, reducing its specificity in oncological imaging [4, 5].

These limitations led to the development of new radiotracers such as radiolabelled choline compounds (e.g. 18F-fluorocholine: N,N-dimethyl-N-[18F]fluoromethyl-2-hydroxy-ethylammonium [FCH]) or amino acid analogues like 18F-fluoro-ethyl-L-tyrosine (FET). Both tracers show a high tumour to background ratio, which facilitates the detection and delineation of especially low-grade and/or cortical tumours. They also seem promising candidates for the differentiation of recurrent brain tumour from radiation injury [6–8]. Moreover, FET, unlike FCH and FDG, does not accumulate in inflammatory cells, a feature which increases the specificity of FET in tumour imaging [9–11].

FET is an amino acid analogue which is not metabolised or incorporated into proteins [12]. A specific Na+-independent amino acid transport system, the L-system, as well as the Na+-dependent system B0, +, is responsible for the high FET accumulation in different tumour cell lines [13, 14]. Wester et al. introduced this tracer to clinical research [12]. Further studies have demonstrated an increased uptake of FET in various brain tumours [8, 15–18], but there are still open questions concerning the main uptake mechanism(s) of FET in vivo. Besides the L- and B0,+ systems, which are located in the membranes of brain endothelial cells (L) and tumour cells (L, B0, +), blood-brain barrier (BBB) disruption, tumour perfusion and the microvessel density (MVD) could influence tracer accumulation. In this respect, Kracht et al. found a positive correlation between the microvessel count and the uptake of [11C]methionine (MET), a radiolabelled amino acid widely used as a tracer in gliomas [19]. To our knowledge no data are available on the correlation between the MVD and FET uptake.

Choline, an important compound of the cell membrane, is transported into mammalian cells by a low-affinity sodium-independent transport system and then phosphorylated by choline kinase [20]. In a further step it is metabolised to phosphatidylcholine, which is incorporated into the cell membrane. In previous studies the increased choline uptake in tumour cells was mainly explained by the upregulation of choline kinase due to an increased demand of membrane constituents [21–23]. However, Henriksen et al. postulated a specific choline transport system in tumour cells as the major mechanism for the clinically important early uptake of radiolabelled choline derivatives [24]. Initial clinical studies with [11C]choline demonstrated the diagnostic potential of this substance in different tumours [25–29]. DeGrado et al. introduced FCH for brain tumour imaging [30]. The major advantage of 18F-labelled compounds, especially in clinical use, is the longer physical half-life of 18F (110 min) compared with 11C (20 min). This allows several PET examinations with one production. The tumour uptake pattern of [11C]choline and 18F-substituted choline analogues is very similar [7]. As mentioned above, tumour uptake seems highly related to transport and choline kinase activity of the tumour cells themselves but can also be influenced by the BBB (specific transport or disruption), tumour perfusion and the MVD.

The aim of this study was threefold: (a) to assess the uptake of the different tracers in the F98 rat glioma, (b) to evaluate the impact of BBB disruption and MVD on tracer uptake and (c) to compare the uptake in the tumours to that in the radiation injuries of our previous study [31]. In that project, radiation injuries were induced with 150 or 250 Gy proton irradiation. Following the development of a circumscribed lesion, the uptake of FET, FCH or FDG was measured using autoradiography and correlated with the histology and the disruption of the BBB.

Materials and methods

Animals

A total of 26 male Fischer 344 rats weighing 250–300 g were used in this study. The experiments were approved by the local veterinary authorities of the Canton of Zurich/Switzerland.

Radiopharmaceuticals

FDG was obtained from the commercial FDG production of the University Hospital Zurich. The production of FET and FCH has been described in detail elsewhere [31].

Cell culture

The rat glioma cell line F98 was provided by G. Mies (Cologne, Germany). F98 cells were grown in DMEM containing 4,500 mg/l D-glucose (GIBCO, Life Technologies, Basel, Switzerland) supplemented with 10% FCS, 2 mM N-acetyl-L-alanyl-L-glutamine (Biochrom AG, Germany) and 20 μg/ml gentamicin (Sigma-Aldrich, Germany). For injection, cells were harvested by trypsinisation, washed 3 times with PBS and resuspended at a final concentration of 50 million cells/ml.

Inoculation procedure

Rats were anaesthetised with an intraperitoneal injection of a ketamine (100 mg/kg; Ketasol 100, Dr. E. Graeub AG, Bern, Switzerland)/xylazine (10 mg/kg; Xylasol, Dr. E. Graeub AG, Bern, Switzerland) mixture and placed in a stereotactic frame. We used the inoculation method of Ambar et al. with minor modifications [32]. After preparation of the inoculation area, a small hole was drilled into the skull. Then, 1×106 F98 cells in 2 μl were implanted into the left parietal cortex at a depth of 1 mm ventral to the dura mater using a 10 μl–26 gauge Hamilton syringe (Hamilton, Bonaduz, Switzerland). After a short incubation time the syringe was slowly removed and the hole was closed. The well-being of the animals were then monitored daily.

Magnetic resonance imaging

Every second rat was examined once with MRI between day 7–14 after inoculation for tumour growth control. Magnetic resonance imaging (MRI) was performed on a 1.5-T system (GE Signa EchoSpeed Plus 1.5 T with Excite II, GE Healthcare, WI, USA). To maximise the signal to noise ratio, animals were positioned in a dedicated wrist coil. The imaging protocol included the following sequences: an unenhanced transaxial T1-weighted spin echo (SE) sequence [repetition time (TR)/echo time (TE)=300 ms/13 ms; slice thickness 3 mm, without an interslice gap] and a transaxial or coronal T2-weighted three-dimensional (3D) fast SE sequence (FSE) (TR/T 3,000/128, slice thickness 1.5 mm). In addition, the T1-weighted sequence was acquired in the transaxial or coronal plane following intravenous administration of gadopentetate dimeglumine (Gd-DPTA; Magnevist, Schering AG, Berlin, Germany; 0.1 mmol/kg body weight) via a tail vein.

Autoradiography with FET, FCH and FDG

After detection of a tumour on MRI scans, autoradiography was performed (9–23 days after inoculation). Under isoflurane inhalation anaesthesia, catheters were placed in the right femoral artery to monitor blood pressure and the femoral vein to allow intravenous administration of the tracers and Evans Blue, which was injected 1 h prior to tracer administration. Evans Blue binds to albumin and therefore acts as an intravascular contrast medium. More information is given in the next section. According to basic biodistribution studies, 15 min (FCH [33] and FET [8]) and 45 min (FDG [34]) were chosen as tracer uptake times. Following injection of 100–150 MBq of tracer, the animals were sacrificed using an overdose of pentobarbital. The brain was removed and instantly frozen in cooled isopentane. For quantification, 10-μm brain slices (100-μm slice distance) were placed on a phosphor imaging screen together with 14C standards and left for 240 min. Tritium-sensitive screens (Fuji TR2025) were used as their uncoated, thin, sensitive layer yields higher resolution 18F autoradiographs than ordinary screens. The data were scanned (Fuji BAS 1800 II, pixel size 50 μm) and converted to kBq/cc. For this conversion the 14C standards had previously been recalibrated using the data of a 4-h exposure of 10-μm slices of a brain homogenate containing a defined amount of 18F activity.

For quantitative analysis the activities were then decay corrected to the time of injection. Dividing these values by the amount of injected activity per gram of body weight yielded standardised uptake values (SUVs). Regions of interest (ROIs) were subsequently placed over the area of average tracer uptake in the tumours and over the contralateral healthy cortex using the software PMOD [35]. The division of tumour by contralateral cortex SUVs revealed tumour to normal brain ratios (TBRs).

Morphological characterisation

For morphological analysis the brains were taken and fixed in 4% formalin/PBS, followed by haematoxylin/eosin (H&E) conventional stain and GFAP (glial fibrillary acidic protein) immunohistochemistry. Briefly, slides were incubated with anti-GFAP antibody (1:1,000, DAKO Z0334) without pretreatment. Slides were then developed using a Ventana machine and the iView DAB (DAKO) development kit.

For histological analysis of the MVD and the BBB leakage, brains were frozen in isopentane cooled to −50°C. For the visualisation of Evans Blue, 10-μm thick slices were fixed with formalin 4% and stained with DAPI (molecular probes, D-1306).In each animal the Evans Blue distribution was evaluated on one typical slice through the tumour. For this purpose the sections were investigated in a fluorescence microscope (Leica MZ16 FA, Leica Microsystems AG, Wetzlar, Germany) with the following filter combination: excitation 540–580 nm, emission 610 nm low pass. The data were stored as an 8 bit intensity image. ROIs were then placed on the tumour and the contralateral cortex, and the mean of the ROI was taken as a measure for relative fluorescence intensity. As was demonstrated in the work by Saria and Lundberg [36], this intensity is linearly correlated with the absolute density of Evans blue. The ratio of the fluorescence intensity in tumour and contralateral cortex was taken as a measure of BBB disruption, which was then correlated with the SUV of tracer uptake.

The MVD was assessed in 22 rats using the von Willebrandt factor stain (vWF). Briefly, brain slices were pretreated with protease 1 for 4 min and stained with polyclonal antibody against vWF (1:1,000, DAKO A00802). The staining was developed using iVIEW DAB (DAKO). The number of vessels was counted in a totally visible tumour diameter of one representative section according to Weidner et al. [37]. For comparison, the counts were then adjusted to number per 10 microscopic high-power fields (0.1885 mm2 per HPF).

Statistics

The present tumour data were compared with those regarding radiation injury from our previous study [31]. The combined data were analysed using a two-way analysis of variance with type of lesion (tumour, radiation injury) and type of tracer (FCH, FET, FDG) as factors. Spearman rank correlation analyses were performed between tracer uptake in tumour tissue (quantified as SUV and TBR) and both microvessel density (MVD) and extravasation of Evans Blue (EB). Owing to the relatively small number of animals, the calculated p values were not corrected for multiple comparisons.

Results

Animals

No animal suffered from systemic side-effects during the period of observation, except for two rats which developed neurological signs. These animals were immediately euthanised.

Development of the F98 glioma

Intracerebral lesions were detected on the MRI scans 7–14 days after inoculation. The diameter of the tumour ranged from 3 to 6 mm. Typical examples of MRI scans are demonstrated in Fig. 1. The tumour was located within the cortex of the left temporo-parietal region and extended into the white matter.

Fig. 1
figure 1

The axial MRI scans of the rat brain clearly depict the F98 glioma in the left temporo-parietal cortex. On the T2-weighted three-dimensional fast SE sequence the tumour demonstrates increased signal intensity (a, arrow). On the T1-weighted SE scan the signal intensity is decreased (b, arrow)

Full size image

Morphological characterisation

Histological slices of a typical F98 glioma infiltration are shown in Fig. 2. The glioma cells diffusely infiltrate the surrounding brain parenchyma. Occasionally, tumour cells formed round and expanding cell bulks. Atypical mitosis were often seen.

Fig. 2
figure 2

Microscopic images of F98 glioma cell transplants in rat brains. a The side of transplantation in the left hemisphere near the corpus callosum (H&E, ×1.625). b Higher magnification of the infiltration zone of the F98 glioma cells (H&E, ×40). c Tumour microvessel (centre) filled with erythrocytes (H&E, ×400). d The perivascular infiltration of the tumour cells (H&E, ×400). Cells spread along microvessels in the Virchow-Robin perivascular space. e Immunohistochemically labelled host astrocytes and F98 glioma cells (anti-GFAP, ×100). As in human glioblastoma, F98 glioma transplants show only occasional positivity for GFAP. f Anti-vWF labelled microvessels within the F98 glioma transplant (anti-vWF, ×200)

Full size image

Disruption of the blood-brain barrier

Typical examples of Evans Blue fluorescent scans are illustrated in the left panels of Fig. 3. Extravasation of Evans Blue as an indication of BBB disruption was present in each case.

Fig. 3
figure 3

The left side of the figure (a, c, e) depicts the disruptionof the BBB as demonstratedby the leakage of Evans Blue. The uptake of FCH (b), FET (d) and FDG (f) is shown on the right

Full size image

Microvessel density

Transplanted tumour cells exhibit a homogeneous tumour mass with a fine network of small arterioles and capillaries. A representative section of the tumour mass shows an evenly distributed microvascular network (Fig. 2f). Summarising, the mean vessel count in the total sample of 22 histologically analysed rats was 179±59 (mean±standard deviation) per 10 HPFs (0.1885 mm2 per HPF).

Tracer uptake and correlation with the extravasation of Evans Blue, microvessel density and uptake in radiation injury

Examples of autoradiographs of each tracer are demonstrated on the right side of Fig. 3. Each tumour displayed markedly higher tracer uptake than the surrounding tissue. The uptake values are summarised in the left part of Table 1. For comparison the corresponding values in the radiation injuries of our previous study [31] are presented on the right of Table 1 and illustrated in Fig. 4. FDG displayed the highest SUV, followed by FET and FCH. The tumour/cortex ratio was highest for FCH. In the tumours the mean SUV was 1.97 (FCH), 2.71 (FET) and 2.37 (FDG) times higher than in the radiation injuries. For FET there was no overlap of the SUV in the tumour and radiation injury, while for FCH and FDG the tumour SUV was in the same range as the uptake in the radiation injury in only one animal. For the lesion to cortex ratio there was no overlap at all. The analysis of variance performed on the SUV data demonstrated significant main effects of lesion type (F (1,39)=45.71, p<0.001) and type of tracer (F (2,39)=25.42; p<0.001) as well as a significant interaction between the two (F (2,39)=8.03; p<0.005). It is important to note that the methodology for the derivation of the SUVs was the same in the gliomas and the radiation injuries.

Table 1 SUVs of F98 gliomas and radiation injury
Full size table
Fig. 4
figure 4

Comparison of tracer uptake in F98 glioma and in radiation injuries from a previous study. For the presentation the data were normalised to the mean of the uptake in the radiation injuries. The data points represent individual animals. The bars are centered on the mean and represent ±1SD. The numbers signify the mean SUV of the original data and the coefficient of variation (SD/mean×100) in percent

Full size image

The correlation of tracer uptake (SUV, TBR), MVD and extravasation of Evans Blue (EB) is demonstrated in Fig. 5. A significant positive correlation was found between relative uptake of FCH and MVD. The same tracer demonstrated a trend towards a negative correlation between uptake and extravasation of Evans Blue. Another trend towards a relevant correlation was present for FET SUV and extravasation of Evans Blue.

Fig. 5
figure 5

Correlations between tracer uptake [standardised uptake value (SUV) and tumour to normal brain ratio (TBR)] and microvessel density (MVD) and the extravasation of Evans Blue (EB). The r values represent Spearman’s correlation coefficient

Full size image

No correlation was found between MVD and extravasation of Evans Blue (data not shown).

Discussion

The F98 rat glioma model proved to be very reliable. Because the tumours have similar characteristics to certain human high-grade brain tumours [38], the results are relevant for the interpretation of human PET scans. For instance, the F98 rat glioma has similar growth characteristics to the glioblastoma multiforme, the most common human primary brain tumour, as is demonstrated in Fig. 2. Furthermore, the cell line is syngeneic in inbred Fischer rats, which leads to a negligible immune response. This is important because tumour uptake is not confounded by uptake into activated inflammatory cells.

Tumour uptake of the different tracers

Not unexpectedly, FDG displayed the highest tumour SUV of all tracers. The lowest SUV was measured with FCH. Although it must be mentioned that a direct comparison of our data with human studies is limited by various factors, such as different acquisition protocols and modalities as well as interspecies differences, for all tracers the SUVs and TBRs were in the range of those observed in human high-grade gliomas [3, 6, 16, 18, 39]. This correspondence is another indication of the usefulness of the used glioma model.

The second aim of the study was to evaluate the effect of BBB disruption and MVD on tracer uptake. This aim was inspired by another study, where the uptake of [11C]methionine in 21 glioma patients was reported to correlate with MVD [19]. In our study no such correlation was found for FET, the amino acid analogue among our investigated tracers. This result suggests that the increased FET uptake in the F98 glioma is mainly determined by a mechanism located in the tumour cells themselves whereas active transport across the BBB into the interstitial space is probably not a rate-limiting factor. Very likely candidates for this mechanism are up-regulated amino acid transporters like system L [L amino acid transporter (LAT) 1–3] and B0,+ in the tumour cell membrane [14]. According to previous studies, the LAT2 and B0,+ systems seem to be the main FET transporters in F98 glioma cells [14, 40].

In addition, we found a trend towards a positive correlation between FET uptake and the extent of BBB leakage, which almost reached significance. This result is in line with our previous experiments in radiation injuries and cryolesions. These experiments demonstrated that a disruption of the BBB led to considerable leakage of FET into interstitial space [31].

In contrast to FET, a significant correlation between tracer uptake and MVD was found for FCH. This finding is in line with the work by Shinoura et al., which demonstrated that [11C]choline TBRs in brain tumours increased with vessel density [41]. In addition, we found a tendency towards a negative correlation between FCH uptake and extravasation of Evans Blue (Fig. 5). This indicates that the choline carriers which transport FCH across the vessel wall into the interstitial space are a relevant factor in the kinetic chain determining FCH uptake [6, 42]. A higher degree of BBB disruption is probably associated with less function of the choline carriers in the vessel wall, which might explain the lower FCH uptake with increasing BBB disruption. A possible passive leakage of FCH across a disrupted BBB does not seem to compensate for the decreased function of the choline carriers in tumour vessels. The present study does not elucidate the importance of the mechanisms located in the tumour cells for FCH accumulation. However, there is evidence that a choline-specific transport system is primarily responsible for the early uptake of radiolabelled choline analogues in different tumour cell lines [24]. Further studies should also address the role of choline kinase on FCH accumulation using a specific inhibitor like MN58b [43].

For FDG no significant correlation of tracer uptake and MVD or BBB disruption was found. This is in contrast to the human study by Aronen et al. [44]. These authors correlated the FDG uptake in 21 gliomas with microvascular blood volume (TBV) in the tumours measured with MRI. They reported a relevant positive correlation in 16 of these tumours. The fact that Aronen et al. used different types of glioma may explain the discrepancy with our study. It is known that higher grade of malignancy is often coupled with increased MVD [37, 45]. However, experimental proof is lacking on whether the increased MVD directly leads to higher tracer uptake. It is well possible that an altered mechanism in the malignant cells is responsible for the increased tracer retention and that the higher MVD is coincidental. In fact, our study with a single cell line favours this latter possibility for FDG and FET. For FCH, the MVD seems to play a more direct role in tracer retention. This may be relevant if the tracer is to be used for tumour grading. The potential for such grading has been demonstrated for FDG [1, 46, 47]. Since increased MVD is associated with higher tumour malignancy and seems also to be a factor determining FCH uptake, the latter might in addition be related to tumour malignancy. This possibility should be tested in further studies. In contrast, FET PET does not seem to be a reliable diagnostic tool for grading brain tumours [16, 48–50]. However, this tracer is very suitable for the differentiation of malignant from benign lesions [8, 9, 11, 15, 16, 18].

Comparison of tracer uptake in F98 gliomas and radiation injuries

The differentiation of radiation injury from recurrent tumour is a relevant clinical problem in brain tumour patients who have been irradiated. In our previous study we investigated the uptake of all three tracers in experimentally induced acute radiation injuries in healthy rat brains [31]. The comparison clearly demonstrated that the uptake of all used tracers was significantly higher in the F98 gliomas (Table 1, Fig. 4). For SUV or lesion/contralateral cortex ratios there was no or only minimal overlap between the tumour and the radiation injury group. The significant interaction of lesion type and tracer in the analysis of variance indicates that there is a difference in the ability of the investigated tracers to distinguish tumour from radiation injury. Visual inspection of Fig. 4 reveals that FET seems most favourable. However, FCH and FDG also seem very suitable.

In clinical practice, the situation is more complicated. FDG has several drawbacks in the differentiation of radiation injury from recurrent tumour. For instance, FDG uptake varies over a wide range in different brain tumours, and some low- and even some high-grade tumours display the same or lower FDG SUVs than are measured in acute radiation injuries [3, 39]. Another problem is the high accumulation of FDG in normal brain tissue, which often makes it difficult to identify a lesion.

The situation seems more favourable for FET. In addition to the present study, other publications have indicated that FET PET holds promise for the differentiation of benign therapy-induced lesions from recurrent tumour [8, 16]. Gliomas generally display a higher uptake than benign lesions like radiation injuries.

For FCH the available data on human brain tumours are still very limited. The only available SUV data were published by research groups at the Gunma University School of Medicine, Maebashi, Japan, and the University of Turku in Helsinki, Finland, both using [11C]choline PET [3, 51–53]. The SUV was in the range 0.17–4.40 in different high-grade gliomas and 0.07–3.31 in low-grade gliomas. For comparison, SUVs in non-neoplastic brain lesions (n=5) ranged from 0.17 to 1.22, indicating overlap with the tumours. In contrast, our investigations demonstrated that FCH is a promising tracer for the differentiation of tumour from radiation injury.

In summary, the accumulation of FCH, FET and FDG is highly increased in F98 rat gliomas relative to uptake in normal brain. Higher MVD led to higher FCH uptake while it had no effect on the accumulation of FET or FDG. The degree of BBB disruption seems to influence the accumulation of FET and FCH, but not that of FDG. For all tracers, tumour uptake was significantly higher than the accumulation in radiation injuries, with almost no overlap between the groups. This is important for the differentiation of tumour from radiation injuries in clinical applications.