Ex vivo MR spectroscopic measure differentiates tumor from treatment effects in GBM

Abstract

The motivation of this study was to address the urgent clinical problem related to the inability of magnetic resonance (MR) imaging measures to differentiate tumor progression from treatment effects in patients with glioblastoma multiforme (GBM). While contrast enhancement on MR imaging (MRI) is routinely used for assessment of tumor burden, therapy response, and progression-free survival in GBM, it is well known that changes in enhancement following treatment are nonspecific to tumor. To address this issue, the objective of this study was to investigate whether MR spectroscopy can provide improved biomarker surrogates for tumor following treatment. High-resolution metabolic profiles of tissue samples obtained from patients with GBM were directly correlated with their pathological assessment to determine metabolic markers that correspond to pathological indications of tumor or treatment effects. Acquisition of tissue samples with image guidance enabled the association of ex vivo biochemical and pathological properties of the tissue samples with in vivo MR anatomical and structural properties derived from presurgical MR images. Using this approach, we found that metabolic concentration levels of [Myo-inositol/total choline (MCI)] in tissue samples are able to differentiate tumor from nontumor and treatment-induced reactive astrocytosis with high significance (P < .001) in newly diagnosed and recurrent GBM. The MCI index has a sensitivity of 93% to tumor in recurrent GBM and delineates the contribution of cellularity that originates from tumor and astrocytic proliferation following treatment. Low levels of MCI for tumor were associated with a reduced apparent diffusion coefficient and elevated choline-N-acetyl-aspartate index derived from in vivo MR images.

Fig. 1.

Association of HR-MAS spectrum with in vivo MR imaging metrics. Outline of the procedure for obtaining in vivo MR parameters corresponding to tissue sample locations for the image-guided study. BrainLab Software was used to obtain the coordinates of the sample location (circle) at the time of surgery for each individual patient. Since these locations were provided in imaging coordinates, it was possible to derive the corresponding anatomical, diffusion, and spectroscopy parameters at this location. There is a good association between parameters derived from the ex vivo and in vivo MR measures (shaded box). A clear correspondence was seen in the Cho levels between the ex vivo and in vivo studies, although given the poor resolution of 3 T MRS, the composition of elevated Cho as originating from a GPC is not clear. Please note that since the in vivo spectrum was acquired at a long TE = 144 ms, metabolites such as mI were not measured due to their short T2 relaxation times. Hence future studies will develop a short echo MR spectroscopic imaging acquisition to detect mI and evaluate MCI in clinical studies.

Fig. 2.

Comparison of tumor and gliosis/astrocytosis HR-MAS spectra. An ex vivo HR-MAS spectrum obtained from a tumor sample confirmed with pathology as 100% tumor (black) and a sample from epilepsy (red) confirmed to have prominent astrocytosis. Levels of mI are significantly higher in the epilepsy sample, representing astrocytosis.

Fig. 3.

Comparison of MCI and H&E tumor cellularity scores for all samples from patients with newly diagnosed GBM. A statistically significant (P < 0.001) difference in MCI levels between tumor samples (cellularity = 3) and nontumor samples (cellularity = 0 or 1) is observed in samples from patients with newly diagnosed GBM. Levels of MCI measured with HR-MAS spectroscopy differentiate tumor from nonumor regions in newly diagnosed GBM. These findings are consistent with the nonimage-guided study in which tumor samples had lower MCI.

Fig. 4.

Comparison of MCI and astrocytosis score in a patient with recurrent GBM. High magnification digitized GFAP immunostained image of 2 samples from the same patient with recurrent GBM. Both samples demonstrated increased cellularity based on H&E scores. An MCI of 2.8 in (A) represents reactive astrocytosis and is consistent with an astrocytosis score (=3) signifying that reactive astrocytes (arrows) are a primary contributor to cellularity. Conversely, an MCI of 0.9 (B) for tumor is consistent with a region composed predominantly of tumor cells, as indicated by an astrocytosis score of 0. This illustrates that MCI has the potential to characterize the cellularity as originating from reactive astrocytosis or tumor in recurrent GBM.

Fig. 5.

Comparison of astrocytosis scores and MCI for all samples from patients with recurrent GBM. There is a statistically significant (P < .001) difference in MCI between tumor samples with an astrocytosis scores for tumor (=0) and astrocytosis (=3). As in the nonimage-guided study, low values of MCI are associated with tumor regions. Higher values of MCI represent reactive astrocytosis, which is a surrogate for treatment effect. These findings indicate that levels of [Myo-inositol/Total Choline] measured with MRS differentiate tumor from treatment effects in patients with recurrent GBM.

Fig. 6.

Comparison of the MCI with the in vivo MR imaging measures. (A) A significant (P < .001) inverse (r = .6) relationship was found between MCI derived from the HR-MAS of the tissue sample and the CNI obtained from the 3T in vivo MRS voxel containing the location from where the tissue was acquired. Low MCI for tumor is consistent with elevated CNI in these voxels. MCI is consistent with the increased presence of astrocytosis following treatment since there are more samples with elevated MCI in samples obtained from patients with recurrent GBM (solid circles) compared with patients with newly diagnosed GBM (open circles). (B) A significant (P < .001) correlation (r = .52) was found between nADC and MCI. Low values of ADC are associated with tumor, which is consistent with lower values of MCI for tumor in these regions. Elevated ADC is generally associated with edematous or nontumor regions, which is consistent with high MCI for nontumor in these regions.

Fig. 7.

Comparison of MCI with contrast enhancement. MCI was lower in regions with increased nCE in patients with newly diagnosed GBM. This is consistent with regions of contrast enhancement being associated with tumor in newly diagnosed GBM. This was not observed in samples acquired from patients with recurrent GBM. This suggests that enhancement is not an accurate surrogate for tumor presence following treatment, highlighting the need for new tumor markers in recurrent GBM.