Multimodal Elucidation of Choline Metabolism in a Murine Glioma Model Using Magnetic Resonance Spectroscopy and ¹¹C-Choline Positron Emission Tomography

Malignant gliomas are the most common primary brain tumors, which arise in approximately 18,000 people annually in the United States, thereby yielding high levels of morbidity and mortality (1). Even with optimal treatment, the median survival time following the diagnosis of a glioblastoma does not exceed 15 months (2). Current concepts about the origin and initiation of gliomas include a complex network of cellular and molecular interactions (3). Because of an enhanced cell proliferation and consecutive elevated levels of cell membrane synthesis during tumorigenesis as well as abnormal physiologic environments, such as hypoxia and acidic extracellular pH, choline metabolism is also involved in brain cancer (4). Choline is a precursor for the synthesis of the phospholipid components of the cell membrane. Cellular choline is phosphorylated by choline kinase (CK) yielding phosphocholine (PCho), which reacts further with CTP to yield CDP-choline. The de novo synthesis (Kennedy pathway) of phosphatidylcholine (PC) then results from the reaction of CDP-choline with diacylglycerol. Approximately, one half of the total membrane lipid content refers to PC (5). Several protein-mediated systems for choline transport exist: (i) high-affinity choline transporters (CHT), (ii) the choline transporter-like family (CTL), (iii) polyspecific organic cation transporters (OCT), and (iv) organic cation/carnitine transporters (OCTN). In the clinical setting, 2 imaging modalities allow for the examination of choline metabolism, namely proton magnetic resonance spectroscopy (¹H-MRS) and positron emission tomography (PET) using ¹¹C-choline. MRS provides noninvasive measurements of tissue concentrations of metabolites, such as total choline-containing compounds (tCho), total creatinine (tCr), and N-acetylaspartate (NAA). Increased tCho levels result from elevated cell membrane turnover and cellular density (6). PCho and glycerophosphocholine (GPCho) are the major contributors of the tCho signal (7). While some authors recommend tCho/tCr ratios and compare them with the contralateral hemisphere (8), other groups favor tCho/tNAA ratios (choline to NAA index = CNI; ref. 9), which has a high correlation with the histologic grading (10). Doblas and colleagues have mentioned that both NAA and tCr levels can decrease during tumor progression, and therefore ratio variations could be relatively insensitive or even remain constant despite changes in the individual metabolite concentrations (11). ¹¹C-choline and its derivatives have been applied for neuro-oncologic PET imaging in the last years (12, 13). While the level of choline uptake in the normal brain is relatively low, the overexpression of the aforementioned amino acid transporters in tumors results in high tumor-to-background ratios. The modeling and quantification of the kinetic behavior of ¹¹C-choline is a challenging task, and a consentaneous approach has yet to be published. While Hara and colleagues have applied tumor-to-reference area radioactivity concentrations (14), Utriainen and colleagues have used the Patlak method (12). Interestingly, differences between the kinetic and the steady-state method were not observed (12).

Although MRS and ¹¹C-choline PET results reflect the choline metabolism, no statistical significance was found with either method (12). Because of this result, the conflicting data in the literature about reference tissues (e.g., the contralateral healthy brain) and metabolites (e.g., NAA) as well as the need for the validation and cross-correlation of both methods (15), we aimed to conduct a comparative in vivo study assessing tCho levels by MRS and ¹¹C-choline metabolism by PET with morphologic parameters in a murine orthotopic glioma mouse model. Such a cross-correlation of metabolic parameters in vivo is of special interest as an emerging imaging technology, hybrid PET/MRI, just entered the preclinical and clinical field (16, 17).

Murine SMA-560 (spontaneous murine astrocytoma) glioma cells (18) were cultured in Dulbecco's Modified Eagle's Medium containing 10% fetal calf serum, 2 mmol/L glutamine, and penicillin (100 IU/mL)/streptomycin (100 mg/mL). Five thousand SMA-560 cells were stereotactically implanted into the right hemisphere of 8- to 16-week-old male and female syngeneic VM/Dk mice on day 0. The local authorities approved the animal experiments. The PET and magnetic resonance measurements were conducted in 35 VM/Dk mice, some of these animals were studied longitudinally, which resulted in a total of 53 complimentary PET and magnetic resonance datasets. The measurements are grouped according to the postimplantation days (d) 5 to 18 (d5, n = 4; d6, n = 4; d7, n = 4; d8, n = 10; d11, n = 4; d12, n = 8; d14, n = 11; d16, n = 6; and d18, n = 2). Additional PCR data were obtained for 6 VM/Dk mice.

First, ¹¹C-choline was applied with a mean injected dose of 12.3 ± 1.2 MBq followed by dynamic microPET imaging (Inveon dedicated PET, Siemens Healthcare) with a spatial resolution in the reconstructed image of 1.4 mm. The following parameters were used: acquisition time 3,600 seconds, ordered subset expectation maximization reconstruction (16 subsets and 4 iterations), attenuation correction, and last 900 seconds summed frames for static image analysis. Subsequently, MRI was carried out in a 7 T small animal MRI scanner equipped with a 290 mT/m gradient system, a whole body quadrature coil for transmission, and a 4-channel mouse brain local coil for magnetic resonance signal reception (ClinScan, Bruker Biospin MRI). Chemical shift imaging spectroscopy data (3D PRESS-CSI; TR (repetition time) = 1,800 milliseconds; TE (echo time) = 135 milliseconds; matrix size, 16 × 16 × 16; voxel size = 1.3 mm × 1.3 mm × 1.3 mm; 2 averages; acquisition time 31 minutes 1 second) was acquired as follows: d5, n = 2; d6, n = 2; d7, n = 2; d8, n = 4; d11, n = 4; d12, n = 5; d14, n = 6; d16, n = 4; and d18, n = 1. Automatic and manual shimming was conducted for an adjustment volume of 6 mm × 3 mm × 5 mm centered in the brain.

Inveon research workplace (IRW, Version 3.0, Siemens Healthcare) was used for image fusion and data analysis. After PET and magnetic resonance image fusion, the regions of interest (ROI) for brain tumor (tumor), normal contralateral brain (brain), and cerebellum were defined on the basis of the T2-weighted magnetic resonance images. Care was taken to exclude the hyperintense regions for the non-tumor regions. As the magnetic resonance sequences covered only the animal brain, the muscle ROIs were identified on the foreleg musculature using the PET datasets. CSI data processing was conducted using an automatic postprocessing routine jSIPRO (java Spectroscopic Imaging PROcessing, Version 0.9, https://www.sites.google.com/site/jsiprotool) based on LCModel (Version 6.2; ref. 19). Spectroscopic images in DICOM format were generated by jSIPRO and imported into IRW for further evaluation. Choline-containing compounds (tCho, at 3.2 ppm) and NAA (at 2.0 ppm) values were obtained from the anatomic ROIs placed on the corresponding CSI maps. Subsequently, the PET data were evaluated using the same ROIs as was applied for magnetic resonance data analysis (except muscle ROIs). The PET data were normalized to the respective amount of injected radioactivity, which is reported as %ID/mL (percentage injected dose per milliliter). Furthermore, PET, CSI, and magnetic resonance data were referenced to different reference regions, thereby yielding the following ratios: tumor-to-brain ratio, tumor-to-cerebellum ratio, and tumor-to-muscle ratio. Details about the correlation analysis between PET and CSI data in 5 example datasets can be found in the Supplementary Material and Methods.

PCR data were obtained from 3 animals each, sacrificed on days 11 and 15 postimplantation. The detailed PCR methods can be found in the Supplementary Material and Methods.

After sacrificing the animals, the mouse brain tissue was dissected from the cranial cavity and fixed in zinc salt solution for 24 hours (20). The tissue was subsequently dehydrated with an increasing gradient of alcohol and embedded in paraffin. The tissue specimens were cut into 5-μm sections and stained with hematoxylin and eosin (H&E). Immunohistochemistry was carried out using an automated immunostainer (Ventana Medical Systems, Inc.) according to the company's protocols for open procedures with slight modifications. The antibody panel used included Ki-67 (SP6; Thermo Fisher Scientific), CD31 (ab28364; Abcam), cleaved caspase-3 (ASP 175; Cell Signaling Technology), CD3 (SP7; DCS Innovative Diagnostik-Systeme GmbH & Co. KG), and glial fibrillary acidic protein (GFAP; 6F2; Dako Deutschland GmbH). The appropriate positive controls were used to confirm adequate staining. The histologic samples were analyzed by an experienced pathologist (L. Quintanilla-Martinez).

For the visualization of the distribution of choline metabolites, our histologic methods were complemented by secondary ion mass spectrometry—mass spectrometry imaging (SIMS-MSI). For details of the methodology please refer to the Supplementary Materials and Methods.

Descriptive statistics including, the mean and 1 SD values, were calculated with SigmaPlot 11.0 (Systat Software, Inc.). Correlation analysis between CSI and PET values was based on ROIs defined by anatomic magnetic resonance and placed to the respective CSI and PET datasets. The correlation between the pairs of variables was evaluated using the Pearson product-moment coefficient. Statistical significance was considered with P ≤ 0.05.

The mean volumes of the intracranial tumors derived from the T2-weighted images increased exponentially between day 5 (6.93 ± 3.49 mm³) and 18 (107.20 ± 29.13 mm³; Fig. 1A). Interestingly, the strongest increase is noticeable starting from day 11, in which the tumor volume quintupled from 12.25 ± 5.02 to 61.62 ± 29.53 mm³ on day 14.

This trend was also observed with the analysis of ¹¹C-choline uptake in the tumor, which starts to prominently increase from 1.27 ± 0.55%ID/mL on day 12 to 3.93 ± 0.28%ID/mL on day 18 (Fig. 1B). This increase is highly significant (P < 0.001). A first peak in ¹¹C-choline uptake at day 7 and a subsequent decrease until day 12 was observed. Finally, the injected doses increased until day 18. The PET tumor-to-brain and tumor-to-cerebellum ratios are presented in the Fig. 1C and D. More detailed values can be found in the Supplementary Results.

The mean tumor tCho values (in arbitrary units, a.u.) decreased from day 5 (5.35 ± 1.65 × 10⁴ a.u.) until day 7 (2.42 ± 0.11 × 10⁴ a.u.). From day 8 (2.99 ± 1.26 × 10⁴ a.u.) until day 16 (2.94 ± 1.44 × 10⁴ a.u.) a plateau was observed. Day 18 indicates a further decrease (1.78 × 10⁴ a.u.) in tCho levels (Fig. 1E). Tumor-to-brain ratios ranged from 1.20 ± 0.09 on day 5 to 2.08 on day 18 (Fig. 1F), whereas the tumor-to-cerebellum ratios for choline-CSI did not continuously increase with a minimum value at day 6 (0.57 ± 0.22) and a maximum value at day 11 (4.45 ± 2.49). The tumor NAA values decreased from day 5 (15.20 ± 5.13 × 10⁴ a.u.) to day 18 (0.36 × 10⁴ a.u.). The metabolite values (averaged from day 5 until day 18) observed in the brain were 2.72 ± 1.33 × 10⁴ a.u. for tCho and 7.30 ± 4.11 × 10⁴ a.u. for NAA (decreasing from 137.00 ± 38.50 × 10⁴ a.u. (d5) to 4.09 ± 2.52 × 10⁴ a.u. (d12) and decreased further to 1.62 × 10⁴ a.u. on day 18), whereas the values for the cerebellum were 2.86 ± 2.54 × 10⁴ a.u. for tCho and 3.56 ± 3.05 × 10⁴ a.u. for NAA. Figure 2 shows an example of PET (Fig. 2A), T2-weighed magnetic resonance imaging (Fig. 2B), CSI data (Fig. 2C) with tumor (Fig. 2D), and brain spectra (Fig. 2E) of a mouse brain with glioma on postimplantation day 18.

The Supplementary Table 1S in the Supplementary Results summarizes the correlations between the values and ratios derived from PET and CSI analyses from day 5 to 18. Identical ROIs for each animal, based on the anatomic magnetic resonance-volumetry images, were used for the respective CSI and PET correlation analysis. Negative correlations were observed between tCho and ¹¹C-choline PET tumor-to-brain ratio (r = −0.680; P = 0.044), tCho and ¹¹C-choline PET tumor-to-cerebellum ratio (r = −0.812; P = 0.008). Positive correlations were observed between tCho tumor-to-brain ratio and %ID/mL in the tumor (r = 0.839; P = 0.005), tCho tumor-to-brain ratio, and ¹¹C-choline PET tumor-to-brain ratio (r = 0.693; P = 0.038) as well as tCho tumor-to-brain ratio and ¹¹C-choline PET tumor-to-cerebellum-ratio (r = 0.659; P = 0.054). The other comparisons revealed no significant correlations. In addition, no correlations were observed between tCho and %ID/mL, NAA and %ID/mL, tCho/NAA and %ID/mL in the normal brain and cerebellum, respectively.

Typical examples (postimplantation days 12–14) of ¹¹C-choline PET, CSI, and T2-weighted–based tumor ROIs of the maximal values are presented in Fig. 3. A visual and a quantitative voxel-wise correlation of PET and CSI values in the tumor ROI reveal overall a low correlation between these 2 measurements (Fig. 3A–E). There is only a small overlap between tumor volumes as identified by CSI and PET (Fig. 3F and I). Highest PET uptake values are present in the intersection between tumor volume as defined by anatomic magnetic resonance imaging and PET (Fig. 3G). The CSI defined tumor volume reveals only a low PET uptake (Fig. 3G), but CSI choline tumor/brain ratios are largest in this volume as well as its intersection with anatomic or PET based information (Fig. 3H). The Venn diagram summarizes these findings of complementarity between PET and CSI tumor volume information (Fig. 3I).

The mRNA expression levels of CHT1, CTL1-4, OCT1-2, CK-α, CK-β, and PHLD for days 11 and 15 in the tumor, brain, and cerebellum tissues are summarized in Fig. 4. Referenced to the brain tissue, most of these important transporters and enzymes were upregulated but displayed a decrease in expression between day 11 and 15. Additional details can be found in the Supplementary Results.

On the basis of the histologic analysis, the tumors were composed of relatively large cells with hyperchromatic and polymorphic nuclei typically with one or more prominent nucleoli (Fig. 5A and E). All tumors displayed a relatively high mitotic rate and areas of necrosis primarily in the center of the tumors at the later postimplantation days. The levels of proliferation as shown with the Ki-67 antibody (Fig. 5B, F, and I) increased relative to the size of the tumor and days after implantation as well as the amount of vascularization visualized by CD31 staining (Fig. 5J). Although apoptosis increased with the size of the tumors, cell death was primarily observed in the center of the lesion, as shown by assessing the levels of activated caspase-3 expression (Fig. 5K). The stains for GFAP showed a strong specific reaction in glial cells predominantly around the tumors, indicating reactive gliosis against the tumor cells (Fig. 5C and G). The formation of gliosis is increasing from day 14 to 18. The CD3 staining shows a substantial infiltration of T cells on day 18 but not for day 14 (Fig. 5D, H, and L).

Figure 6 presents the distributions of free choline (m/z 104) and PCho (m/z 184) in a mouse control brain and a tumor-bearing brain at day 15.

To date, it is not fully understood whether ¹H-MRS and ¹¹C-choline PET reflect the same or different parts of the choline pathway and whether the imaging parameters are complementary or not. Therefore, we have conducted a small animal imaging study on a syngeneic glioma model using VM/Dk mice with orthotopically implanted SMA-560 cells and subsequently validated the in vivo results with in vitro and ex vivo techniques. In such a syngeneic tumor model, immunodeficiency is not required as is with xenograft models; therefore, the tumor growth and microenvironmental characteristics resemble more closely the clinical conditions.

The levels of choline uptake increased over time, which is reflected by a range of the tumor-to-brain ratio values from 1.12 ± 0.15 for day 5 to 2.22 ± 0.11 for day 18. These results are comparable with the human data reported by Kato and colleagues, which displays a tumor-to-brain ratio of 2.69 ± 2.04 in grade 2 astrocytic tumors. However, both grade 3 and 4 tumors have a higher tumor-to-brain ratio with 4.76 ± 3.04 (grade 3) and 18.35 ± 6.73 (grade 4; ref. 21). Our standardized uptake values (SUV) values ranged from 0.31 to 0.71, which are lower as compared with a C6 glioma rat model (2.00 ± 0.53; ref. 13). The lower values are due to the chemically modified tracer, the different tumor entities, the partial volume effects, and the different weight composition of rats compared with mice. The shape of the tumor %ID/mL curve reveals the first maximum at day 7, which decreased until day 12 and then increased again until day 18 (Fig. 1B). We hypothesize that the shape of the curve is caused by a complex interplay of transporter regulation and partial volume effects. At the initiation of tumor growth, a majority of the transporters are upregulated (see the PCR data for day 11 and Supplementary Discussion), thereby causing an increase in choline uptake. This measured choline uptake in tumors is amplified by a strong partial volume effect in the small tumors during the initial phase. The rapid tumor growth following the initial phase reduces the partial volume effect in combination with a transporter downregulation, which is reflected in a reduction in the %ID/mL values until day 12. As the tumors further progress, the %ID/mL values increased in proportion with the tumor volume. Both the tumor-to-brain and tumor-to-cerebellum ratio curves reflect this trend but exhibit another minimum on day 16 (Fig. 1C and D). The tumor-to-muscle ratio values became increasingly unsteady over time, which is a phenomenon that might be caused by altered muscle metabolism due to cachexia.

The acquired CSI data reflecting the tCho levels in the tumors ranged from 5.35 × 10⁴ a.u. for day 5 to 1.78 × 10⁴ a.u. for day 18. As the CSI acquisitions do not allow for an absolute quantification without further processing, it is difficult to compare our results with other data in the literature. However, the shape of the CSI spectra is qualitatively comparable with those acquired in a clinical setting (Fig. 2D). In the human brain, the choline-containing compounds were found at a concentration of 64.3 ± 10.1 μmol/100 g (human brain), whereas astrocytomas and anaplastic astrocytomas display concentrations of choline-containing compounds of 68.6 ± 6.5 μmol/100 g and 132.6 ± 7.9 μmol/100 g, respectively (22). On the basis of the tumor-to-brain ratios, the values are 1.07 for astrocytoma and 2.06 for anaplastic astrocytoma. These values correspond with our glioma model, in which the tumor-to-brain ratios range from 1.20 for day 5 to 2.08 for day 18. Interestingly, Usenius and colleagues have displayed a decrease in the absolute concentrations of choline in astrocytoma grade 4 (1.62 ± 0.28 mmol/L) compared with the control (1.74 ± 0.09 mmol/L), astrocytoma grade 1–2 (1.93 ± 0.09 mmol/L) and astrocytoma grade 3 (1.95 mmol/L), whereas the PC/(PC + GPC) ratio increased (23). This decrease toward the later tumor stages is also reflected by our data. However, also a decrease in the contralateral hemisphere is present resulting in an increase of the tumor-to-brain ratios. Possible explanations for this phenomenon could be different choline concentrations in the brain (24), accelerated aging (25, 26) and glioma-induced neurodegenerative processes (27), reduced global cerebral blood flow (28), and ischemia (29). Doblas and colleagues have also observed significant variations and heterogeneity in different rodent gliomas, such as the F98 model (8.1 ± 12.4; ref. 11). Therefore, we conclude that our MRS and PET data are in accordance with the corresponding human and small animal data in the literature. The increase of tumor volume from day 5 through 18 was paralleled by increased contrast enhancement (days 8–16) as well as progression of proliferation and vascular density confirmed by histology.

In this study, we observed negative correlations between the tCho and ¹¹C-choline PET tumor-to-brain ratio (r = −0.680; P = 0.044) and the tCho and ¹¹C-choline PET tumor-to-cerebellum ratio (r = −0.812; P = 0.008). Because of the decreasing tCho levels in the brain as a reference tissue (see Supplementary Discussion), positive correlations were observed between the tCho tumor-to-brain ratio and %ID/mL (r = 0.839; P = 0.005), the tCho tumor-to-brain ratio and the ¹¹C-choline PET tumor-to-brain ratio (r = 0.693; P = 0.038). These correlations were obtained when including all the data in the analysis. Because of the significant aggressiveness of the tumor, the group sizes at the last postimplantation days were small. The exclusion of the day 18 data from the analysis impaired the statistical significance of the correlations as follows: the tCho and ¹¹C-choline PET tumor-to-cerebellum ratio (r = −0.754; P = 0.031), the tCho and ¹¹C-choline PET tumor-to-brain ratio (r = −0.537; P = 0.170), and the tCho tumor-to-brain ratio and %ID/mL (r = −0.555; P = 0.153). In opposition to our findings, Utriainen and colleagues have failed to observe a significant correlation between the tCho levels determined by ¹H-MRS and ¹¹C-choline uptake (r = 0.50–0.54; P = 0.1; ref. 12). They have declared that it is uncertain whether the association should be expected because the choline-containing component of MRS represents intracellular metabolite pools of PCho and GPCho, whereas the rate of choline uptake is thought to be controlled by amino acid transporter expression and attenuation in tumor endothelial cells (12). However, in this study, only 8 patients had malignant brain tumors and there was a time lag of a few days between the 2 imaging studies. Both the small sample size and the rapid tumor proliferation might have changed within a few days, which may explain their findings. Rommel and colleagues have observed a correlation between choline levels and ¹⁸F-FCH uptake ranging from 0.12 to 0.33 with P values ranging from 0.054 to 0.48 (30). They concluded that high choline levels correlated with low values for ¹⁸F-FCH uptake and vice versa.

Kato and colleagues have displayed a positive correlation between ¹¹C-choline tracer uptake and the proliferation index in astrocytic tumors (r = 0.64; P < 0.001; ref. 21), which is supported by our data. Shimizu and colleagues have reported a strong linear relationship between the Ki-67–labeling index and the MRS tCho levels (r = 0.81; P < 0.0001) and a weak correlation between the tCho/NAA ratio (r = 0.60; P < 0.02; ref. 31). Our data suggest a negative correlation between tCho levels and the proliferation index; although, when tCho levels were referenced to either the brain or cerebellum, we observed a positive correlation (see Supplementary Discussion).

We also observe complementary data with ¹¹C-choline PET and CSI results in the qualitative outline of the tumor presented as maximal values. CSI highlights the areas of high choline concentration, which seems to be localized to the tumor rim, whereas ¹¹C-choline PET identifies the regions of high choline turnover. Tumor spread to normal brain tissue seems to be better characterized by CSI, which indicates also high tCho levels in the vicinity of the tumor, not observed in ¹¹C-choline PET in these areas. Rand and colleagues documented white blood cell infiltrates in most of the false positive lesions analyzed with MRS (32). The authors hypothesized that these infiltrates are a source of elevated choline. In addition, Srinivasan and colleagues found that gliosis in response to surgical trauma and therapeutic modalities may be indistinguishable from tumor growth (33). Using GFAP and CD3 immunohistochemistry, we could disclose that gliosis and inflammation are increasingly found in the later tumor stage of our murine model. Therefore, the CSI tCho signal at the tumor periphery is potentially linked to inflammation and gliosis, whereas the ¹¹C-choline PET uptake in the tumor core is linked to proliferative tumor cells. These results are in the line with metabolic imaging derived from SIMS-MSI. Here, the free choline is predominantly located in the tumor rim, whereas PCho can be found in the tumor periphery yielding a high tCho CSI signal. On the other hand, the specific ¹¹C-choline PET tracer accumulates more in the proliferative tumor parts.

Our study has several limitations. Because of the significant aggressiveness and the rapid growth rates of the tumors, the animals were susceptible to prolonged imaging times. Because of the animal protection and ethical concerns, animals had to be sacrificed as soon as they exhibited impairments caused by the tumor. This resulted in heterogeneous group sizes. The high morbidity of the animals did not allow for an acquisition of the full image parameters for PET and magnetic resonance. PET and magnetic resonance measurements were conducted sequentially; therefore, small offsets in coregistration must be expected. However, as the brain is stabilized by the skull, coregistration errors between PET and magnetic resonance measurements are on the order of less than 1 mm. As both the PET and CSI voxel sizes are above this limit, we concluded that these errors are negligible. We were unable to conduct an absolute quantification of the choline CSI data because the inclusion of a reference sample would have prevented the use of a tight fitting local coil on the brain of the animal, which would have influenced the signal-to-noise ratio and magnetic resonance image quality. Furthermore, an additional magnetic resonance CSI scan without water suppression was not conducted because this would have significantly extended the measurement time per animal. Given the increased rate of morbidity of the animals at later stages of tumor growth, we aimed to both reduce the time during which the animals were under anesthesia as much as possible and conduct imaging protocols that are feasible in a clinical setting, which are also streamlined for patient comfort and throughput.

In this study, we monitored tumor progression in a mouse glioma model. We found that the magnetic resonance CSI data of choline metabolism were in agreement with the PET ¹¹C-choline data when a reference region, such as the normal brain, was used for MRS quantification. It is important to take into account that the detection limit for MRS techniques is in the 10⁻³ to 10⁻⁶ mol/L range, whereas the PET method concentration limit is 10⁻¹² mol/L. Furthermore, we found that the ¹¹C-choline PET data of brain tumors correlated with both PCR and histologic analyses. Absolute quantification of ¹¹C-choline PET data was more reliable compared with the MRS technique. Therefore, we suggest that these methods have complimentary roles, with MRS functioning as a tool for qualitative tumor assessment, early indication of tumor spread, gliosis and inflammation, whereas ¹¹C-choline PET serves as an imaging biomarker for proliferation, and is therefore of interest for treatment planning and therapy monitoring.

Bernd Pichler received grant/research support from Siemens, AstraZeneca, Bayer Healthcare, Boehringer-Ingelheim, Oncodesign, Merck, Bruker, and the Werner Siemens-Foundation. No potential conflicts of interest were disclosed by the other authors.

Development of methodology: H.F. Wehrl, A.W. Sauter, F. Jiru, R.M.A. Heeren, B.J. Pichler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.F. Wehrl, A.W. Sauter, G. Reischl, K. Hasenbach, F. Cay, D. Bukala, L. Quintanilla-Martinez, G. Tabatabai, A. Kiss

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.F. Wehrl, J. Schwab, A.W. Sauter, K. Chughtai, A. Kiss

Writing, review, and/or revision of the manuscript: H.F. Wehrl, J. Schwab, K. Hasenbach, G. Reischl, G. Tabatabai, L. Quintanilla-Martinez, F. Jiru, K. Chughtai, A. Kiss, F. Cay, D. Bukala, R.M.A. Heeren, B.J. Pichler, A.W. Sauter

Study supervision: H.F. Wehrl, A.W. Sauter

The authors thank Maren Koenig and Mareike Lehnhoff for the excellent technical support and Ursula Kohlhofer for help with the histology during this project. We also appreciate the development and support of the jSIPRO tool by Antonin Skoch and Milan Hajek from the Institute for Clinical and Experimental Medicine.

This study was supported by German Research Foundation (DFG PI 771/1-1, SFB 773), the Werner Siemens-Foundation, and Ministry of Health Czech Republic [grant: 00023001 Institute for Clinical and Experimental Medicine (IKEM), Prague]. SIMS work is part of the research program of the “Foundation for Fundamental Research on Matter (FOM),” financially supported by the “the Netherlands Organization for Scientific Research (NWO).”

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 with 18 U.S.C. Section 1734 solely to indicate this fact.