**3. Biological significance of MRI variables in GBM**

gliomas. Radiolabeled 2-[18F]fluoro-2-deoxy-D-glucose ([18F]FDG), methyl-[11C]-L-methio‐ nine ([11C]MET) and 3-deoxy-3-[18F]fluoro-L-thymidine ([18F]FLT) are taken up by prolifer‐ ating gliomas depending on their tumor grade as the consequence of an increased activity of membrane transporters for glucose ([18F]FDG), amino acids ([11C]MET), and nucleosides ([18F]FLT) as well as increased expression of cellular hexokinase ([18F]FDG) and thymidine kinase ([18F]FLT) genes, which specifically phosphorylate [18F]FDG and [18F]FLT, respec‐ tively [20]. Imaging of brain tumors with [18F]FDG was the first oncologic application of PET. [18F]FDG is actively transported across the blood-brain barrier (BBB) into the brain where it is phosphorylated and trapped into cells. Since 1982 [5,21], PET with [18F]FDG has been accepted and widely used in the grading of brain tumors; its uptake is generally high in highgrade tumors and it has a good prognostic value, because increased intra-tumoral glucose consumption correlates with tumor grade [22], biological aggressiveness and survival of patients in both primary and recurrent gliomas. Pathology and survival can be predicted by [18F]FDG-PET in gliomas [6]. In addition, a tumor-to-white matter ratio and tumor-to-gray matter ratio were found to increase the sensitivity of the grading evaluation [22]. Since intratumor heterogeneity of brain tumors is not adequately revealed in conventional MRI, because evaluation of the contrast enhancing lesion can either under-or overestimate the presence of active tumor, MRSI and PET are requested to gain additional information on metabolic and molecular tumor markers. In a tumor, the grading can be heterogenous with low-and highgrade areas, as it happens frequently in GBM. This may affect the choice of the site for stereotactic biopsy, which must direct towards tumor sites with the highest tumor grade. Therefore, suitable targets for biopsy will have positive contrast enhancement on T1-weighted MRI, a high choline-peak on MRSI and hypermetabolism on [18F]FDG-PET, the uptake of which is much higher in high-grade component of tumors. As a matter of fact, the [18F]FDG-PET improved the diagnostic yield of stereotactic biopsies by detecting metabolically active

70 Tumors of the Central Nervous System – Primary and Secondary

However, [18F]FDG-PET can have some diagnostic limitations, because of the high rate of physiologic glucose metabolism in normal brain tissue. In the brain cortex it is particularly high [24,25], so when a hypermetabolic lesion is close to the cortex or the subcortical white matter, the distinction of the tumor from the normal tissue may be difficult [22]. Moreover, it must be taken into account that [18F]FDG accumulation can be non-specific, because it is also observed in inflammatory or granulation tissues [26]. A later PET image acquisition [27] and a co-registration of PET images with MR images greatly improves the performance of [18F]FDG-PET [28]. Technologic advances have allowed to merge PET and MR images, combining the high resolution of MR imaging with the low-resolution functional capability of PET [23], defined as a reduction of intracellular oxygen pressure (pO2), because of decreased supply and of increased demand for oxygen. It predicts poor treatment response of malignant tumors. Two different forms of tumor hypoxia are recognized. Diffusion-limited chronic hypoxia may develop as a result of increased intercapillary distances, and acute hypoxia can result from occlusion of large tumor vessels [29]. Both forms of hypoxia have several implica‐ tions for the further evolution of tumors (induction of signaling cascades that promote angiogenesis, growth, and cell migration) [30]. Tumor hypoxia may also lead to necrosis, which is mandatory to establish the diagnosis of GBM. The **([18F]Fluoromisonidazole) ([18F]FMI‐**

areas of tumor [23].

Basically, three conditions can be detected with anatomy-based MRI: iso-hypo-intensity in TC1 (tumor, edema), hyper-intensity in TC2 (edema) and contrast enhancement (malignancy). The contrast enhancing regions (CERs) of untreated GBM correspond to the most histologically malignant areas of the tumor with architectural disruption, high cell density, proliferation, vessel density and angiogenesis with circumscribed necroses. Many other properties are revealed by physiology-based MRI. In CERs, in comparison with non-enhancing regions (NERs), physiologic MRI variables show higher values of rT1C, relative fast spin echo (rFSE), rCBV, relative peak height (rPH) and relative recovery factor (rRF), whereas rADC, relative fractional anisotropy (rFA) and fluid attenuated inversion recovery (rFLAIR) do not differ from NERs. All these observations have been shown and confirmed in recent studies of many cases of GBM planning pre-operatively tissue sampling sites and marking them on the anatomic images used by the surgical navigation work station. A comparison between MRI variables and histology of the samples corresponding to the MRI regions of interest (ROIs) in CERs and NERs has been made [32,33]. A correlation of histopathologic features and DWI and DSC variables prevailed in enhancing areas and rCBV and Cho/NAA index (CNI) above and rADC below a certain value could indicate the occurrence of tumor cells. Neoangiogenesis could be recognized and distinguished from simple endothelial hyperplasia, even though the permeability of the region is limited. Interestingly, also T2 rFSE and rFLAIR hyperintensity areas could show histopathologic features of malignancy. On the whole, it was demonstrated by gene microarray that the genetic expression patterns between CERs and NERs were different, with genes associated with mitosis, angiogenesis and apoptosis clustered in CER surgical samples [32].

occurrence or not of tumor cells. In particular, Epidermal Growth Factor Receptor (EGFR) amplification, the occurrence of EGFRvIII and Tumor Protein p53 (TP53) mutations were more frequent in CERs than in NERs, corresponding to a malignant histology. The genetic variability in the different samples was interpreted as due to polyclonality and not only to a genetic heterogeneity supported by the occurrence in the same tumor of different non-tumor cells of various species. Polyclonality means cell complexity, formed by tumor cells that differ among themselves for a series of phenotypic and molecular characteristics of cell proliferation, invasion, *etc.* [41,42]. This observation can be a warning against the use of small tumor samples to characterize the genotype of the entire tumor. Heterogeneity has been explained either by the hierarchic model mechanism [43] or by the stochastic mechanism [44] of tumor develop‐ ment. The two models, however, cannot be mutually exclusive, because their cells should derive from a common ancestor [45]. As for EGFR amplification, the possibility that its variation could depend on an asymmetrical distribution during mitoses must be mentioned [46,47], even if it is already included in the clone formation. The neurosphere assay produces neurospheres (NS), characterized by stemness antigens (Figure 9A,C,E), and adherent cells

Spatial Relationships of MR Imaging and Positron Emission Tomography with Phenotype, Genotype and...

http://dx.doi.org/10.5772/58391

73

However, their phenotype is not the same in the different tumor sites, differing for the quota of the two types of antigens. There must be a different capacity of the tumor areas to host or to generate GSCs and this is in line with the concept that a GSCs hierarchy exists for their potential [48-52]. GSCs have been interpreted as the top of a hierarchy of tumor cells for stemness and, therefore, for self-renewing, clonogenicity, *etc.* They occur in tumor niches and are under the control of microenviroments with their intrinsic and extrinsic signaling [53,54]. The niches can be perivascular or perinecrotic [53] and for a series of observations and considerations they must develop in the most malignant sites of the tumor [51,55]. Stemness and differentiation are the opposite poles of a spectrum in which a hierarchy exists of GSCs as for their potential [48-52]. Going from areas of the highest malignancy, such those of CERs, to differentiated ones, the stem-cell potential decreases [55]. In this way the distribution of GCS in GBM could be explained. The NS and AC degree of differentiation or stemness, demon‐ strated by the relevant antigens, represents an interesting subject of study that has been

Confocal microscopy is an advanced technique of optical imaging used to obtain high resolution images [56,57]. In tissue and cells derived from GBM, it is possible to distinguish the emission signal of different markers and to perform study of both co-localization and quantification of the luminous signal related to the protein marker expression. The cellular heterogeneity is a hallmark of GBM. Using differentiation and stemness markers it is possible to identify hypothetical immature or dedifferentiated elements in the whole tumor cell population, as well as in NS or AC by the neurosphere assay. Confocal images of glioma cells by double immunofluorescence allow to distinguish the expression pattern of the markers above mentioned. Their expression levels, related to the intensity of the emitted signal, show variable Nestin and glial fibrillary acidic protein (GFAP) positivity, depending on the tumor site. The method has a paramount importance in the study of the spectrum from stemness to

(AC), characterized by differentiation antigens (Figure 9B,D,F).

pursued by us by confocal microscopy (unpublished data).

differentiation.

The values of MRSI variables such as Cho/Cr and Cho/NAA showed a parallel variation as those of DWI and PWI. In spite of the possibility of a misregistration between biopsy sites and MRI uploaded to the neuronavigation device if a brain shift occurs, GBM histologic features could be usefully identified by physiology-based MRI [33].

Using the same technique, *i.e.* combining physiology-based MRI, MRSI and [18F]FDG-PET imaging with neuronavigation work station in a series of gliomas, mainly GBMs, we observed that the values of rCBV, ADC, Cho/Cr, CNI were useful for recognizing tumor areas and their phenotypic variations, as for both the number of cells and vascular pathologic structures (Figures 6,7). A possible source of error was the mismatch between the MRI registration and the sampling by the navigator, so that a dissociation between the variable values of the ROIs and histopathology occurred. For example, a ROI on central necrosis could erroneously correspond to a high rCBV value and histologically to the occurrence of tumor tissue in the sample. However, this was a rare event and it did not prevent from recognizing the biological significance of the imaging values contained in the ROIs, also by extrapolation among all the samples.

Malignant gliomas are hypermetabolic in comparison with normal brain. The glycolytic metabolism is increased as well as protein and membrane synthesis to maintain the rapidly dividing tumor cells. MRSI in spite of an intra-or inter-subject variability can identify the tumor. There are two patterns clearly distinct: one is that corresponding to the ROIs on necrotic regions and the other that on the enhancing ring. In the first one, two patterns have been in turn described: "necrosis" and "cystic necrosis" with variable Ch and high LL peak and with no peaks and variable LL, respectively. The ROIs on the ring show a high Cho and Cho/NAA ratio, whereas very variable are those on regions around the ring [34]. In our series Cho, Cho/ Cr and Cho/NAA values were constantly high in CERs in comparison with NERs.

Fusing MRI and [11C]MET-PET it was shown that the volume of the radio-compound uptake is greater than that of gadolinium enhancement on T1, even though smaller than T2 volume; it extends beyond in most cases [35,36] correlating with the proliferation markers [37,38], increased Cho/NAA and DTI abnormalities in the white matter [28,39]. However, there could be an underestimation of the tumor extension, because infiltrating cells do not proliferate [3,40].

The number of genetic alterations decreased from the most malignant areas of the tumor to the peripheral areas, correlating fairly well with the MRI variable values and indicating the occurrence or not of tumor cells. In particular, Epidermal Growth Factor Receptor (EGFR) amplification, the occurrence of EGFRvIII and Tumor Protein p53 (TP53) mutations were more frequent in CERs than in NERs, corresponding to a malignant histology. The genetic variability in the different samples was interpreted as due to polyclonality and not only to a genetic heterogeneity supported by the occurrence in the same tumor of different non-tumor cells of various species. Polyclonality means cell complexity, formed by tumor cells that differ among themselves for a series of phenotypic and molecular characteristics of cell proliferation, invasion, *etc.* [41,42]. This observation can be a warning against the use of small tumor samples to characterize the genotype of the entire tumor. Heterogeneity has been explained either by the hierarchic model mechanism [43] or by the stochastic mechanism [44] of tumor develop‐ ment. The two models, however, cannot be mutually exclusive, because their cells should derive from a common ancestor [45]. As for EGFR amplification, the possibility that its variation could depend on an asymmetrical distribution during mitoses must be mentioned [46,47], even if it is already included in the clone formation. The neurosphere assay produces neurospheres (NS), characterized by stemness antigens (Figure 9A,C,E), and adherent cells (AC), characterized by differentiation antigens (Figure 9B,D,F).

CERs and NERs has been made [32,33]. A correlation of histopathologic features and DWI and DSC variables prevailed in enhancing areas and rCBV and Cho/NAA index (CNI) above and rADC below a certain value could indicate the occurrence of tumor cells. Neoangiogenesis could be recognized and distinguished from simple endothelial hyperplasia, even though the permeability of the region is limited. Interestingly, also T2 rFSE and rFLAIR hyperintensity areas could show histopathologic features of malignancy. On the whole, it was demonstrated by gene microarray that the genetic expression patterns between CERs and NERs were different, with genes associated with mitosis, angiogenesis and apoptosis clustered in CER

The values of MRSI variables such as Cho/Cr and Cho/NAA showed a parallel variation as those of DWI and PWI. In spite of the possibility of a misregistration between biopsy sites and MRI uploaded to the neuronavigation device if a brain shift occurs, GBM histologic features

Using the same technique, *i.e.* combining physiology-based MRI, MRSI and [18F]FDG-PET imaging with neuronavigation work station in a series of gliomas, mainly GBMs, we observed that the values of rCBV, ADC, Cho/Cr, CNI were useful for recognizing tumor areas and their phenotypic variations, as for both the number of cells and vascular pathologic structures (Figures 6,7). A possible source of error was the mismatch between the MRI registration and the sampling by the navigator, so that a dissociation between the variable values of the ROIs and histopathology occurred. For example, a ROI on central necrosis could erroneously correspond to a high rCBV value and histologically to the occurrence of tumor tissue in the sample. However, this was a rare event and it did not prevent from recognizing the biological significance of the imaging values contained in the ROIs, also by extrapolation among all the

Malignant gliomas are hypermetabolic in comparison with normal brain. The glycolytic metabolism is increased as well as protein and membrane synthesis to maintain the rapidly dividing tumor cells. MRSI in spite of an intra-or inter-subject variability can identify the tumor. There are two patterns clearly distinct: one is that corresponding to the ROIs on necrotic regions and the other that on the enhancing ring. In the first one, two patterns have been in turn described: "necrosis" and "cystic necrosis" with variable Ch and high LL peak and with no peaks and variable LL, respectively. The ROIs on the ring show a high Cho and Cho/NAA ratio, whereas very variable are those on regions around the ring [34]. In our series Cho, Cho/

Fusing MRI and [11C]MET-PET it was shown that the volume of the radio-compound uptake is greater than that of gadolinium enhancement on T1, even though smaller than T2 volume; it extends beyond in most cases [35,36] correlating with the proliferation markers [37,38], increased Cho/NAA and DTI abnormalities in the white matter [28,39]. However, there could be an underestimation of the tumor extension, because infiltrating cells do not

The number of genetic alterations decreased from the most malignant areas of the tumor to the peripheral areas, correlating fairly well with the MRI variable values and indicating the

Cr and Cho/NAA values were constantly high in CERs in comparison with NERs.

could be usefully identified by physiology-based MRI [33].

72 Tumors of the Central Nervous System – Primary and Secondary

surgical samples [32].

samples.

proliferate [3,40].

However, their phenotype is not the same in the different tumor sites, differing for the quota of the two types of antigens. There must be a different capacity of the tumor areas to host or to generate GSCs and this is in line with the concept that a GSCs hierarchy exists for their potential [48-52]. GSCs have been interpreted as the top of a hierarchy of tumor cells for stemness and, therefore, for self-renewing, clonogenicity, *etc.* They occur in tumor niches and are under the control of microenviroments with their intrinsic and extrinsic signaling [53,54]. The niches can be perivascular or perinecrotic [53] and for a series of observations and considerations they must develop in the most malignant sites of the tumor [51,55]. Stemness and differentiation are the opposite poles of a spectrum in which a hierarchy exists of GSCs as for their potential [48-52]. Going from areas of the highest malignancy, such those of CERs, to differentiated ones, the stem-cell potential decreases [55]. In this way the distribution of GCS in GBM could be explained. The NS and AC degree of differentiation or stemness, demon‐ strated by the relevant antigens, represents an interesting subject of study that has been pursued by us by confocal microscopy (unpublished data).

Confocal microscopy is an advanced technique of optical imaging used to obtain high resolution images [56,57]. In tissue and cells derived from GBM, it is possible to distinguish the emission signal of different markers and to perform study of both co-localization and quantification of the luminous signal related to the protein marker expression. The cellular heterogeneity is a hallmark of GBM. Using differentiation and stemness markers it is possible to identify hypothetical immature or dedifferentiated elements in the whole tumor cell population, as well as in NS or AC by the neurosphere assay. Confocal images of glioma cells by double immunofluorescence allow to distinguish the expression pattern of the markers above mentioned. Their expression levels, related to the intensity of the emitted signal, show variable Nestin and glial fibrillary acidic protein (GFAP) positivity, depending on the tumor site. The method has a paramount importance in the study of the spectrum from stemness to differentiation.

tumors relapses within 2-cm margin around the enhanced region [58]; another adverse characteristic of infiltrating cells is that they do not proliferate [3,40], escaping thus detection

Spatial Relationships of MR Imaging and Positron Emission Tomography with Phenotype, Genotype and...

http://dx.doi.org/10.5772/58391

75

Tumor invasion is not uniform along its borders. It can be non-existent where the tumor sharply abuts against the normal nervous tissue (Figure 10A), as it may happen with the white matter, or it gradually progresses from the tumor outwards (Figure 10B). Typical is the invasion of the cortex from a tumor located in the white matter, even with the typical picture of perineuronal satellitosis (Figure 10C). The different invasion modalities have been codified [59] and a distinction between diffuse and local tumors has even been proposed [60], but it

**Figure 10.** A – Sharp tumor border. Ki-67/MIB.1, DAB, 100x; B – Invasion gradient toward the cortex. H&E, 100x; C –

Letting aside the mechanisms of tumor cell migration and invasion of which there is today a good knowledge [61-63], some information about neuropathologic findings on peritumor tissue are relevant. First of all, it has been demonstrated that patients with absence of tumor cells in the adjacent normal nervous tissue had better survival than those with tumor cells [64]. This is not in contrast with the observation that the removal of edematous peritumor tissue

Perineuronal satellitosis. H&E, 400x; D – Infiltration along corpus callosum. H&E, 200x.

was not confirmed by the observation of substantially different outcomes.

and being less sensitive to treatments.

**Figure 9.** Immunofluorescence (IF) for stemness antigens in NS. A – Nestin, 200x; C – CD133, 200x; E – Musashi1, 200x. IF for differentiation antigens in AC. B – GFAP, 400x; D – GalC, 400x; F – β-III Tubulin, 400x. Nuclei are counterstained with DAPI.
