The motivation of this study was to address the urgent clinical problem related to the inability of MR imaging measures to differentiate tumor progression from treatment effects in patients with GBM. Contrast enhancement on anatomical MR imaging has proven to be an inadequate surrogate of tumor presence primarily because it has not been linked with biological mechanisms of tumor growth and response. Its lack of association with the pathogenic condition is worsened following treatment. Antiangiogenic and cytostatic therapies affect the morphological and physiological characteristics of the tumor that impact observable MR intensities, such as contrast enhancement, in complex ways that are as yet not well characterized or understood. Hence contrast enhancement following such therapies does not provide an accurate assessment of tumor burden or progression-free survival. The goal of this study was to investigate whether MRS can provide improved biomarker surrogates for tumor following treatment.
A tumor region is characterized by cellular proliferation due to rapidly dividing tumor cells. The MRS total Cho level has been associated with increased cell membrane turnover and shown to correlate with cellular density.19
Maximum Cho levels were shown to differentiate20
low- and high-grade brain tumors with proton MR spectroscopic imaging (1
. Reactive astrocytosis is a normal cellular reaction to tumor growth and treatment. Reactive astrocytes, macrophages, and activated microglia populate regions adjacent to an aggressively infiltrating tumor mass and are commonly dispersed within the “reactive stroma” of a treated tumor. The presence of reactive astrocytosis may therefore be a biomarker surrogate for treatment effects and contribute significantly to cellularity following treatment. High concentrations of mI have been found in astrocytes and were related to activation of glial cells in vitro.16
Free intracellular mI is an osmolyte21
and during periods of osmotic stress, balance is preserved by regulation of mI transport across the plasma membrane.22
Many diseases that are characterized by astrocytic proliferation and demyelinization such as Alzheimer's disease23
and multiple sclerosis24
are associated with increases in mI. It has been argued25
that mI might be used as an indicator of response of normal tissue to tumor infiltration. Studies26
have shown a trend toward lower mI levels in anaplastic astrocytomas and GBMs compared with low-grade astrocytomas. Given these findings, it is anticipated that inclusion of mI would be important for developing a diagnostic tool for differentiating tumor from treatment effects in GBM.
The approach in this study was to use pathological assessment as a gold standard to confirm the presence of tumor or treatment effects and find metabolic markers that are consistent with this assessment. While studies have shown the potential of MRS measures as biomarkers for tumor pathogenesis, a correlative pathological assessment as in this study is important to validate the tumor biology represented by metabolic markers. Using epilepsy samples in the nonimage-guided study enabled the investigation to be focused primarily on observable metabolic differences likely to represent reactive astrocytosis and not confounded by tumor pathogenesis. Correlating MRS-pathological parameters in paired image-guided tissue samples enabled a direct association of biochemical and morphological assessment of tumor tissue. The nonimage-guided study provided an independent data set to derive MRS concentration cutoffs that represent tumor and reactive astrocytosis. The findings in this study were consistent with known biochemical mechanisms of tumor pathogenesis. MRS studies have demonstrated relationships between Cho, PC, GPC, Gly and tumor grade and biologic aggressiveness.27–29
Increase in alanine in tumor samples can be linked with its role in tumorigenesis in GBMs.30
Cho kinase activity has been implicated in the pathogenesis of several tumors and has a role in the transformation of normal to a malignant phenotype.31
Relative levels of PC and GPC enable discrimination of low and high-grade glioma.32
The Cho kinase enzymes are implicated in the synthesis of PE, which was shown to be elevated in transformed cells or tumor samples using31
. Cr are reflective of cellular energy metabolism and are decreased in tumor relative to normal tissue. While mI itself provides marginal differentiation between astrocytosis and tumor, constructing the ratio of mI with total Cho to generate MCI highlights this distinction because it normalizes the effect of astrocytosis by the total cellularity and represents the dominant contribution of a region that could have both tumor and astrocytosis. While the ability of mI to represent astrogliosis has been suggested in prior studies,25
to our knowledge this is first study to perform direct pathological-MRS correlative studies to demonstrate this relationship.
The MRS-pathological correlative studies in this manuscript are in excellent concurrence. Overall, the findings of the image-guided study are consistent with the nonimage-guided study and provide further confirmation of MCI levels to differentiate tumor from nontumor (Fig. ) in newly diagnosed GBM and reactive astrocytosis (Fig. ) in recurrent GBM. Figure illustrates that MCI levels are able to characterize the increased cellularity in a recurrent tumor region as originating from reactive astrocytosis or tumor cells. Both the samples in Fig. were acquired from the same recurrent GBM patient, and the diverse properties of the samples highlight the significant biological heterogeneity of the tumor region in these patients. MCI demonstrated high sensitivity (~93%) for tumor in recurrent GBM, which is remarkable considering the small sample size, heterogeneity of GBM and treatments, and the fact that only a fraction of samples had scores for both pathology and MRS due to analysis thresholds for quality checks. The average MCI for nontumor or astrocytosis samples (Figs and 5) was more than 2.5 times or greater than 2 standard deviations of the MCI for tumor samples in these analyses. This illustrates that the differentiation of tumor and nontumor using this index can be made with strong statistical significance and confidence for patients with GBM. In both image-guided and nonimage-guided data sets, total Cho levels were not found to differentiate between tumor and reactive astrocytosis. This supports our hypothesis that since both tumor and reactive astrocytic proliferation can result in increased cellularity they are indistinguishable by total Cho measurements that represent cell density. While the components that contribute to the Cho peak, that is, GPC or PC, may be different in tumor and astrocytosis due to low resolution of an in vivo MRSI spectrum at 1.5 and 3.0 T, only total Cho measurements can be made in clinical studies. Hence total Cho in in vivo MRSI may not be able to differentiate tumor from treatments effects. An inherently lower level of mI in GBMs compared with low-grade astrocytomas26,34
and an increased presence of reactive astrocytosis from normal brain processes following treatment further contribute to the sensitivity of MCI in GBM.
The concurrence of MCI to in vivo measures of tumor in the image-guided study further strengthened its role as a potential marker to differentiate tumor from nontumor and astrocytosis. 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 (Fig. A: solid circles) compared with patients with newly diagnosed GBM (Fig. A: open circles). Regions with relatively high ADC are generally associated with edema or nontumor and are consistent with elevated MCI, representing astrocytosis in these regions (Fig. B). Despite the strong correlation between CNI and MCI (Fig. A), it is noted that MCI provides an independent characterization of the tumor region. As constructed, CNI in an in vivo spectrum is driven by the extent of neuro-axonal loss (NAA) in a tumor region. In regions of extensive neuro-axonal loss due to tumor growth and significant production of GPC and PC by viable tumor cells that contribute to total Cho, CNI is a strong indicator of tumor presence, and these regions will also have a low MCI. However, in regions that have moderate presence of NAA such as residual tumor regions following treatment, MCI is able to differentiate residual tumor from reactive astrocytosis. Since residual tumor regions are of primary interest in evaluating tumor progression or progression-free survival, the ability of MCI to differentiate these regions will be important for the evaluation of therapy response. These findings also suggest that biochemical properties measured with MRS maintain their sensitivity to tumor metabolism following targeted and focal therapies and provide improved surrogates for tumor presence following treatment.
Since the in vivo MR spectroscopic measure of CNI has been used to locate tumor regions in GBM,17
during the course of the analyses a direct comparison of the ability of MCI and CNI in the same tumor sample to predict pathological assessment was attempted. However, a significant proportion of the samples did not have any NAA, which was consistent with image-guided tissue acquisition primarily from tumor regions guided by anatomic, diffusion, perfusion and spectroscopic imaging definitions of tumor presence. Hence, it was difficult to compare the relative abilities of CNI and MCI in predicting tumor from treatment effects in GBM. Future short-echo MR spectroscopic imaging studies will provide mI, NAA, and Cho. This will allow a prospective evaluation of tissue samples acquired from regions of presumed tumor and reactive astrocytosis based on CNI and MCI and enable a direct comparison of these indices in predicting these pathological states. Figure illustrates that MCI follows the anticipated relationship with contrast enhancement in newly diagnosed GBM. Low MCI for tumor is associated with increased enhancement and elevated MCI for nontumor is found in samples originating in regions with decreased enhancement. However, this relationship between MCI and contrast enhancement is not found in samples from recurrent GBM. For untreated newly diagnosed GBM, contrast enhancement is a fairly good indicator of tumor presence since it is the result of disruption of the blood–brain barrier from tumor growth. Hence low MCI for tumor correlates well with increased contrast enhancement in newly diagnosed GBM. Patients with recurrent GBM have undergone surgery, radiation, and/or antiangiogenic therapies. Posttreatment enhancement in these patients can result from disruption of vasculature from these treatments and not exclusively from tumor presence.35
This results in a weaker relationship of MCI with contrast enhancement in recurrent GBM and highlights the need for alternative surrogates for tumor presence following treatment.
The potential for making an in vivo measurement of the MCI enhances the clinical relevance of this index. Robust measurements of mI36
in association with other metabolites, such as Cho, Cr, and NAA, can be made at 3 T with short-echo MR spectroscopic imaging (MRSI). This MRSI acquisition can be integrated in clinical research MR protocols for patients with GBM undergoing treatment to enable a prospective evaluation of MCI in an independent image-guided data set. The ability to provide a noninvasive, accurate tumor classification, particularly for recurrent GBM without having to perform tissue acquisition and pathological analysis is very powerful for clinical diagnoses and evaluation of disease progression in longitudinal studies. Optimizing the in vivo acquisition of MCI, prospective evaluation of MCI, and correlation of in vivo MCI to clinical outcome to ascertain the prognostic value and biological relevance of this MR spectroscopic index will be investigated in future studies.