[1] 10w The Glucose Transporter and Blood-Brain Barrier of Human Brain Tumors

[1] 165w All grades of astrocytic tumors were studied to assess how variations in the state of differentiation and relative density of astroglial cells affected glucose transporter expression. Only the highest-grade areas of anaplastic astrocytomas and glioblastomas were considered, since histological diagnosis is based on the highest-grade area observed. In addition, a small sam- Therefore, despite 2 "positive" specimens, no glioblastoma stained with the completeness and intensity of low-grade astrocytomas or the majority of anaplastic astrocytomas. These results are summarized in Figure 3. Because of the heterogeneous morphology of glial tumor vessels, attention was paid to the staining characteristics of specific foci of vascular proliferation (coiled masses of small vessels, often with enlarged endothelial cells [14]). Although the incidence of vascular proliferation did increase with increasing grade, this abnormal morphology never accounted for the majority of vessels. The staining of these proliferative vessels by glucose transporter antiserum was consistent with that of the tumor in toto (being either positive, negative, or mixed) in all but 1 specimen.

[2] 48w The presence of contrast enhancement on preoperative imapine s m d i ~s wm taken as evidence of loss of the permeability barrier that normally limits interendothelial diffusion. Variations in vascular permeability among the grades of tumors allowed us to assess the relationship between permeability and transporter expression.

[3] 26w Two of 9 low-grade astrocytomas demonstrated contrast enhancement preoperatively. Seven of 9 anaplastic astrocytomas showed enhancement. All 10 glioblastomas and all 3 metastases showed contrast enhancement.

[4] 176w The frequency of glucose transporter positivity in relation to that of contrast enhancement for each grade of glial tumor is represented in Figure 4. Only specimens for which imaging studies were available were included in the analysis. Low-grade astrocytomas and glioblastomas demonstrated an inverse relationship between transporter expression and contrast enhancement. However, most anaplastic astrocytomas were transporter-positive despite an abnormally high vascular permeabdity evidenced by contrast-enhanced imaging. This association occurred in 6 of 9 anaplastic astrocytomas. Only 1 of 9 low-grade astrocytomas and 2 of 10 glioblastomas stained positively and showed contrast enhancement. Remaining tumors were either transporter-positive and nonenhancing (7 low-grade astrocytomas and 1 anaplastic astrocytoma), transporternegative and enhancing (1 anaplastic astrocytoma, 6 glioblastornas, and 3 metastases), or mixed and enhancing (1 low-grade astrocytoma and 2 glioblastomas). reaction product. Vessels of the low-grade astrocytoma and anaplastic astrocytom stain positively, while metastasis and gliobhstoma vessels are negative for glucose transJorter. UEA I and FVIIIrAg stain all vessels. As expected, ePythmytes are also stained by glucose transporter antiserum. (Hemtoxylin One anaplastic astrocytoma was both transporternegative and nonenhancing.

[1] 73w The blood-brain barrier is more than a simple restriction of Permeability. Specialized enzymes and transport systems regulate the entry of substances into the brain [2). These proteins form a biochemical component of the barrier, which is necessary for normal brain function. In pathological conditions, barrier function is usually disturbed but not eliminated [ 181. However, most assessments of blood-brain barrier function include only permeability studies, both in the laboratory and in the clinic.

[2] 63w We assessed a biochemical property of the bloodbrain barrier in human astrocytic neoplasms, a pathological condition in which the permeability barrier has been extensively studied. Using immunohistochemical techniques, we showed that blood vessels of such tumors may express apparently normal glucose transporter levels. This high expression was seen in lowgrade and anaplastic astrocytomas, but the expression of transporter decreased significantly in glioblastoma multiforme.

[3] 146w We also investigated the relationship between glucose transporter expression and contrast enhancement, the clinical measure of barrier permeability. It has been shown that the vessels of normal, nonenhancing cerebrum express high glucose transporter levels [4, 5}. If the expression of transporter and permeability restrictions are strictly linked, one would expect vessels of nonenhancing tumors to be transporterpositive. This association held for 8 of 7 such tumors. Conversely, vessels of enhancing tumors would be expected to be transporter-negative or to stain with a mixed pattern. This was only true for 13 of 22 enhanc- ing tumors. Anaplastic astrocytomas accounted for the majority of tumors that were transporter-positive despite contrast enhancement (see Fig 4). In fact, as anaplasia progressed, permeability increased before transporter expression decreased. The expression of glucose transporter independently of permeability restrictions suggests that these two barrier components may be induced or maintained by separate mechanisms.

[4] 91w Similar observations have been made in other systems. In muscle, capillary endothelial cells express higher levels of both the muscle and brain-type glucose transporter than do surrounding myocytes (17-j. This high expression occurs despite the ready permeability of these vessels to horseradish peroxidase {20]. In the developing rat brain, levels of glucose transporter pro- tein and messenger RNA (mRNA) both decline transiently just after birth, a time when permeability restrictions are increasing 1213. Hence, in normal muscle and developing rat brain, microvascular glucose transporter expression is regulated independently of vessel permeability.

[5] 153w The heterogeneity of astrocytic neoplasms makes it possible that sampling error accounts for the identification of transporter-positive yet contrast-enhancing tumors. Pathological specimens staining positively might be derived from nonenhancing regions within an enhancing tumor. However, studies correlating CT and MR images with histological findings consistently show that in lesions demonstrating contrast enhancement, solid tumor was found only in enhancing areas and not in nonenhancing regions 122-241. The latter areas contained only isolated tumor cells infiltrating intact parenchyma. Since this study assessed only specimens of solid tumor tissue and excluded infiltrated border zones, it is very likely that transporterpositive specimens from enhancing lesions represented areas of increased vascular permeability. It is possible but unlikely that rare negative vessels, undetected in our specimens, account for enhancement (e.g., by allowing escape of contrast, which diffuses over a large area). Our use of two vascular markers to detect all vessels within a specimen should minimize this error.

[6] 200w Normal brain parenchyma can induce many biochemical and structural features of the blood-brain barrier, even in vessels that normally do not express such a barrier 125, 261. Of the many brain cells potentially responsible, attention has long been focused on the astrocyte 1271. Astrocytic foot processes ensheathe 99% of brain capillaries, sharing a common basement membrane to form an astrocyte-endothelial complex 1281. Developmentally, this astrocytic investment is temporally associated with capillary maturation {29} and the appearance of biochemical barrier components [30,31). Janzer and Raff recently showed that pure astrocytes transplanted to the chorioallantois are vascularized by vessels impermeable to Evans blue 1103. Endothelial cells cocultured with astrocytes demonstrate increases in the number and complexity of tight junctions 11 1). Astrocytes also produce plateletderived growth factor and fibroblast growth factor 132, 331, both of which are known to induce the glucose transporter in fibroblasts 1343. C6 glioma cells, which express astrocytic properties 135, 361, are capable of inducing gamma-glutamyl transpeptidase 13 71 and polarization of amino acid transport [383 in cultured endothelium, both of which are barrier-associated properties. Of particular interest, endothelial cells increase their glucose uptake in the presence of either C6 cells or normal astrocytes, but not oligodendrocytes 1131.

[7] 249w Tumors caricature properties of normal tissue, and despite their altered behavior, unmask normal biological processes {39]. In astrocytic tumors, vessels are bathed in an environment abnormally enriched in astroglial-derived cells, and relatively depleted of other cell types. Our observation that low-grade astrocytomas express normal transporter levels, along with the known inductive capacity of astrocytes, suggests that astrocytes induce glucose transporter expression in human brain. Ultrastructural studies revealed an essentially normal capillary ensheathement by neoplastic cells in these well-differentiated tumors 18, 93, preserving the intimate astrocytic-endothelial relationship. Expression of transporter was maintained in more anaplastic tumors and was consistently absent only in glioblastoma multiforme, whose vessels are incompletely invested by extremely dedifferentiated cells 17, 93. Although this study does not exclude the possibility that nonastroglial cells persisting in these tumors are the actual inducers, this is unlikely because of the extremely low density of such cells relative to normal and the disruption of their normal cytoarchitectural relationships. It is also unlikely that glucose transporter induced by another cell type is merely retained by vessels growing into astrocytomas. Metastatic tumors did not express transporter despite a brain-derived vascular supply, and multiple reports of peripheral tissue transplants into brain consistently failed to demonstrate vessels retaining barrier-associated properties 125, 261. However, it is difficult to prove that specific inductive influences occur simply by studying brain tumor sections. Therefore, we are currently investigating an in vitro model of neural angiogenesis to assess more definitively the role of astroglial cells in glucose transporter induction 140).

[8] 115w Our findings may have implications for the therapy and prognosis of astroglial neoplasms. It is unknown whether glucose flux into tumors with permeable vessels is significantly increased by transporter expression. If, despite contrast enhancement, the presence of specialized transporters signifies that glucose diffusion into parenchyma is restricted, then entry of chemotherapeutic agents might also be limited. In such tumors, novel systems for drug delivery may be more efficacious 141). The ability of a tumor to induce high transporter levels may indicate a retention of differentiated properties associated with lower relative malignancy and improved prognosis. If so, the presence of glucose transporter by immunostaining may provide a means of subclassifying astroglial tumors relevant to patient survival.

[1] 160w Recently resected adult supratentorial tumors were chosen for study. Preference was given to lobectomy specimens, because they usually included normal brain tissue that could serve as a positive control during immunohistochemical staining. Specimens were fixed in 10% neutral buffered formdin for 6 to 18 hours and then processed to paraffin blocks. The microscopic slides from all specimens were reviewed and the neoplasms categorized into one of three grades: low-grade astrocytoma, anaplastic astrocytoma, or glioblastoma multiforme, according to Burger and Vogel [ 14). Briefly, low-grade astrocytomas were remarkable for mild to moderate hypercellularity of uneven distribution and mild nuclear pleomorphism with occasional mitoses. Anaplastic astrocytomas were characterized by marked hypercellularity and nuclear pleomorphism. The diagnosis of glioblastoma multiforme was made only when coagulative necrosis involving the neoplastic astrocytes was identified [15}. The presence of vascular proliferation was not used as a criterion for grading. Ten low-grade astrocytomas, 10 anaplastic astrocytomas, 10 glioblastomas, and 3 metastases (all primary lung carcinomas) were studied.

[2] 236w Serial 5-pm-thick sections were deparaffinized, rehydrated, and stained with antiserum to glucose transporter, Ulrx europaeus I (UEA I) lectin, or antibody to Factor VIIIrelated antigen (FVIIIrAg). The glucose transporter antiserum is a rabbit antiserum to the carboxy terminus of the human erythrocyte transporter, developed and characterized by one of us (L. R. D.) [5]. This antibody and similar antibodies produced by other groups have been shown to selectively stain vessels expressing a permeability barrier in normal brain C4, 51. Rehydrated sections were incubated for 15 minutes at room temperature with 3'3% hydrogen peroxide (H202) in 0.05 M Tris-buffered normal saline solution, pH 7.6, and then for 10 minutes with 2% bovine serum albumin (BSA, Sigma Chemical, St. Louis, MO) in 0.5 M Tris buffer, pH 7.6. Sections were then incubated overnight at 4°C with 1 : 1,000 glucose transporter antiserum. Controls consisted of substituting nonimmune rabbit serum for immune serum. Sections were then incubated at room temperature in 1 : 100 biotinylated goat anti-rabbit immunoglobulin (Vector, Burlingame, CA) for 30 minutes, followed by l . 100 horseradish peroxidase-streptavidin conjugate (Dako, Santa Barbara, CA) for 30 minutes, and developed with 3-amino-9-ethylcarbazole chromogen (Biomeda, Foster City, CA) for 10 minutes. Sections were counterstained with hematoxylin. Dilutions were made in 0.5 M Tris buffer containing 1% BSA, p H 7.6, unless noted otherwise, and slides were rinsed between incubations with 0.05 M Tris-buffered normal saline solution, p H 7.6.

[3] 148w Stains for endothelial cell markers were performed as just described, with the following modifications. UEA I lectin staining was done with 1 : 1,000 UEA I lectin (Vector) and 1 : 2,000 biotinylated goat anti-UEA I (Vector) substituted for primary and secondary antibodies, respectively. Sections for staining with anti-FVIIIrAg were digested with Pronase (Calbiochem, San Diego, CA), 1 mg/ml, in 0.05 M Trisbuffered normal saline solution containing 0.2 96 ethylene diaminetetraacetic acid (EDTA) for 30 minutes at 37°C just prior to incubation with 1 : 20 normal horse serum (Cappel, Cochranville, PA) in place of 2% BSA. Primary and secondary antibodies were I . 50 mouse monoclonal anti-FVIIIrAg (Dako) and 1 : 100 biotinylated horse anti-mouse (Vector). Controls for vascular markers consisted of substituting an irrelevant antibody for UEA I lectin or normal mouse serum for WIIIrAg. All controls performed were negative for glucose transporter and vascular markers.

[4] 43w For each specimen, preoperative imaging studies were assessed for the presence or absence of contrast enhancement as a marker of the permeability barrier. Magnetic resonance (MR) scans and computed tomograms (CT) were considered equivalent, and in no case where both were available did

[5] 33w The Table lists characteristics of the specimens studied. Specimens of glioblastoma and metastasis were from older patients, as expected. The high incidence of temporal lobe tumors reflects the selection of lobectom y specimens.

[1] 171w The glucose transporter of the human brain has been localized to endothelial cells expressing the blood-brain barrier, but little is known regarding its mechanism of induction or whether its expression is exclusively linked with restricted vascular permeability. We investigated glucose transporter expression by vessels in human astrocytic tumors and pulmonary metastases to the brain using immunohistochemical techniques. Vessels in 9 of 10 low-grade astrocytomas and 8 of 10 anaplastic astrocytomas were positive for glucose transporter. Glioblastoma vessels were transporterpositive in only 2 of 10 specimens. Vessels in all three metastatic tumors were negative for the glucose transporter. The decrease in transporter expression observed in higher-grade tumors occurred independently of increases in vascular permeability. In low-grade astrocytomas and glioblastomas transporter expression and contrast enhancement were inversely related, but vessels in 6 of 9 anaplastic astrocytomas were transporter-positive despite contrast enhancement. These findings suggest thit separate mechanisms induce the glucose transporter and the permeability restrictions of the human blood-brain barrier. They also have potential implications for the therapy and prognosis of astroglial neoplasms.

[2] 25w Guerin C, Laterra J, Hruban RH, Brem H, Drewes LR, Goldstein GW. The glucose transporter and blood-brain barrier of human brain tumors. Ann Neurol 1990;28:758-765

[3] 77w The blood-brain barrier is composed of a complex array of physical, metabolic, and transport properties located at the capillary endotheliurn. Structural endothelial specializations, consisting of tight intercellular junctions, an absence of fenestrations, and few cytoplasmic vesicles, impose a permeability barrier limiting the diffusion of polar compounds El}. In addition, a collection of endothelial transporter proteins and enzymes serves to regulate delivery of important substances to the brain and comprises the biochemical component of the blood-brain barrier [2].

[4] 109w Under normal conditions, the brain derives its energy h o s t entirely from glucose oxidation. In the face of permeability restrictions imposed by the bloodbrain barrier, a large and constant glucose supply is maintained by specialized facilitative transporters located at the endothelium f2, 31. Since microvessels account for less than 0.1 % of brain weight, and every molecule of glucose used must cross the endothelial cell, the density of such transporters on endothelia must be considerably higher than on surrounding neurons and glia. Using quantitative analyses, Kalaria and colleagues showed the glucose transporter concentration on human brain microvessels to be among the highest of any tissue studied 141.

[5] 106w Recently, using immunohistochemistry and cytochalasin B-binding assays, the glucose transporter of the human brain was localized almost exclusively to vessels expressing a permeability barrier [4, 51. Whereas cerebral and cerebellar microvessels stain intensely, those of the area postrema and adenohypophysis are noc stained. Neuropil is stained minimally or not at all. Ultrastructurally, immunoelectron microscopic examination revealed that the transporter is indeed located at the endothelial cell plasma membrane 151. Given these findings and the brain's dependence on glucose, it may be postulated that expression of high glucose transporter levels by cerebral microvessels is tightly linked to the presence of the permeability barrier which limits interendothelial diffusion.

[6] 121w The metabolism and vasculature of human brain tumors, especially those of the astrocytic series, have been studied extensively. These tumors may use up to three times as much glucose as normal brain {b]. Their capillaries have a range of morphologies, from essentially normal in low-grade tumors to markedly abnormal in higher-grade tumors 17-91. Abnormally high vascular permeability is demonstrated by contrastenhanced imaging studies. One might logically suspect that the need for specific glucose transporters is de-creased or eliminated in tumors with permeable vessels since glucose should passively diffuse into brain parenchyma, as do the much larger proteins. Alternatively, increases in vascular permeability might not be associated with decreased transporter expression if the mechanisms that establish these two barrier properties are distinct.

[7] 49w Many blood-brain barrier properties can be induced in vitro by astrocytes 110-1 31, but glucose transporter induction has not been addressed. In astrocytic tumors, astroglial cells markedly predominate. Therefore, glucose transporter expression by vessels of astrocytic tumors would suggest a role for the astrocyte in transporter induction in vivo.

[8] 27w To clarify these issues, we used the diverse characteristics of human astrocytic tumors to explore the relationships between glucose transporter expression, blood-brain barrier permeability, and astroglial-derived cells.

[9] 66w Endothelial markers were used to assess the vascular density of each specimen, allowing an estimate of the fraction of vessels positive for the glucose transporter. These markers have been shown to identify vessels in astrocytic tumors as efficiently as those of normal brain C16, 17). For anaplastic astrocytomas and glioblastomas, only the highest-grade areas were considered, since histological diagnosis is based on the hghest-grade area observed.

[10] 100w Tumors were placed into three categories with regard to struning by antibody to the glucose transporter: "positive" when all observed vessels were stamed (1 anaplastic astrocytoma and 1 glioblastoma with only small, focal areas of negative vessels were included in this category); "mixed" when significant numbers of stained vessels and unstained vessels were seen; and "negative" when all observed vessels were unstained (2 anaplastic astrocytomas and 1 glioblastoma with only small, focal areas of positive vessels were included in this category). The intensity of staining relative to that found in normal brain tissue was noted for positive and mixed tumors.

[11] 119w Age Sex Side Tumor n (yr; mean 3-SEM) W F ) (RIL) Lobe Low-grade astrocytoma 10 50 +-4 416 218 5 T , 4 F , 1 P Anaplastic astrocytoma 10 44 * 4 713 317 3 T, 3 F, 2 P, 1 FIP, 1 deep Glioblastoma multiforme 10 59 r 6 614 317 5 T , 2 P, 1 F, 1 PIO, 1 deep Metastases 3 56 4 013 2/1 1 T, 1 F, 1 FIP T = temporal; F = frontal; P = parietal, 0 = occipital. results disagree. If actual films were unavailable, official final reports were used (6 cases). Only noncontrast imaging was performed preoperatively in 1 case of low-grade astrocytoma and 1 anaplastic asrrocytoma.