B. L. Coomber 1, P. A. Stewart 2, K. Hayakawa 3, C. L. Farrell 4 and R. E Del Maestro 5. 1,2,3Department of Anatomy, University of Toronto, Toronto, Ontario, ...
Journal of Neuro-Oncology 5: 299-307 (1987) © Martinus Nijhoff Publishers, Boston - Printed in the Netherlands
299
Quantitative morphology of human glioblastoma multiforme microvessels: structural basis of blood-brain barrier defect B. L. C o o m b e r 1, P. A. Stewart 2, K. Hayakawa 3, C. L. Farrell 4 and R. E Del Maestro 5
1,2,3Department of Anatomy, University of Toronto, Toronto, Ontario, Canada M5S 1A8 (1current address. Vascular Research Laboratory, Toronto General Hospital Research Centre, CCRW 1-857, 200 Elizabeth St., Toronto, Ontario, Canada)," 4,SBrain Research Laboratory, Victoria Hospital, 375 South St., London, Ontario, Canada N6A 4G5
Key words: blood-brain barrier, glioblastoma multiforme, endothelial ultrastructure, capillary permeability
Summary Neoplastic invasion of the brain parenchyma results in a disruption of the ultrastructure of the blood vessel walls such that serum proteins extravasate into the surrounding tissue, resulting in cerebral edema. The structural changes involved are not well understood, since the pores through which serum constituents pass (permeability routes) in normal barrier vessels and in tumor vessels where the barrier is compromised, have not been extensively explored. In this study we investigate the ultrastructure of h u m a n brain microvessels in biopsied samples of control brain tissue and five glioblastoma multiforme tumors. Electron micrographs of a total of 78 vessels were analysed with computer assisted m o r p h o m e t r y for ultrastructural evidence of permeability routes. Fenestrations in the endothelium were not seen. Pinocytotic vesicle number and arrangement did not differ significantly from that seen in control brain vessels. Interendothelial junctions with enlarged distensions (which may represent sections through transendothelial channels) were seen in some vessels from most tumors but not in control barrier vessels. In addition, large gaps in the endothelial layer were seen in less than two percent of t u m o r vessels. In conclusion, glioblastoma multiforme vessels in this study show subtle alterations in vessel m o r p h o l o g y from that seen in controls. We suggest that the high vascular permeability and resultant brain edema seen in glioblastoma multiforme tumors is likely due to the presence of channels through interendothelial junctions, and rare but large breaks in the endothelial wall.
Introduction The vasculature of brain tumors arises from the normal surrounding brain vessels. Blood vessels already present in the normal invaded tissue, as well as those induced to penetrate the tumor, are probably subjected to some additional unknown influence from the t u m o r cells that causes the blood vessels to alter their structure and function, loosing their barrier characteristics. Permeability of the vascular wall is hypothesized to be due to the movement of solutes through water
filled channels [1, 2]. The pathway available to water and small solutes is thought to be represented by narrow channels called junctional clefts through interendothelial junctions that can be seen in the electron microscope as areas where the outer leaflets of the adjacent membranes are separated by a narrow space [3]. In fenestrated endothelium, an additional route for small solutes and water likely exists through channels present in fenestral diaphragms [4]. The pathway for large molecules as well as water and small solutes likely exists in two forms, one coupled to hydraulic flow (i.e. through a water filled
300 channel, probably tubulo-vesicular structures) and the other via some route independent of water movement (possibly vesicle shuttling) [5]. When the barrier breaks down, as it does in a variety of neuropathological conditions, the permeability routes involved may differ from pathology to pathology. It is not clear whether permeability routes vary with tumor type, as descriptive evidence suggests [6-8]. The most common primary brain tumor (i.e. composed of transformed ceils of neuroepithelial origin) is glioblastoma multiforme. Qualitative descriptions of vessel ultrastructure from these tumors describe junctional abnormalities, increases in vesicle numbers, and rare attenuated regions of endothelial cytoplasm that appear fenestrated [9, 10]. Other workers, however, report that fenestrations are rare in human glioma vasculature [8, 11]. The study reported here quantitatively examines tumor endothelium for morphological characteristics held in common with known 'leaky' vessels, and compares it to normal barrier vessel endothelium, to determine which permeability route(s) may be affecting the development of vasogenic edema in this tumor type.
Methods
Source of tumors Samples of glioblastoma multiforme tumors were obtained during lobectomy for treatment of the malignancy. Patients selected for this study were all diagnosed as having primary glial tumors, with no other known disease that might affect microvessel ultrastructure. All patients were normotensive, and had not had prior radiation treatment or chemotherapy. All patients received Haxadrol phosphate (Organon Ltd., Canada), four i.v. doses of 0.25 mg/kg body weight and mannitol, 1.0 g/kg body weight in 20°70 solution infused i.v. 15 minutes prior to surgery. Experimental samples for this study were removed from the brain and immediately placed into 2.5070 buffered glutaraldehyde, pH 7.3, and minced into very fine pieces. In addition, samples of 'control brain', taken from as far away from the tumor as possible, were fixed by the same method. Pathologi-
cal examination of the samples confirmed the diagnosis that the tumors were glioblastoma multiforme. These tumors are intensely cellular and are composed of neoplastic astrocytes with hyperchromatic and pleomorphic nuclei. Mitotic activity is enhanced. Some pseudopallisading of cells was seen, as were necrotic areas with perivascular sparing of cells. Tumors were located in either frontal, temporal or occipital lobes of the cerebrum ('normal' neocortical microvessels do not differ morphologically from lobe to lobe; [12]), and were obtained from five patients (three male and two female), ranging in age from 22 to 77 years.
Tissue processing Tissue samples remained in glutaraldehyde for 4 to 6 hours, then were rinsed in phosphate buffer and osmicated for one and one half hours with 1% Os 04. Blocks were further rinsed in buffer, stained en bloc with uranyl acetate, dehydrated in ethanol followed by propylene oxide and embedded in E P O N and Araldite epoxy resins. Sections of approximately 1/zm in thickness were stained with toluidine blue and examined under the light microscope. Blocks that contained areas of necrotic tissue were eliminated from the study, as these were lacking in patent blood vessels. Such necrotic areas with no blood flow do not contribute to edema formation. Although glioblastoma multiforme tumors are very heterogeneous, vessel morphology is reasonably uniform. We choose blocks that were entirely composed of tumor cells and vessels, as we were not interested in examining vessels from peritumoral regions in this study. Blocks were then trimmed for further sectioning for electron microscopy. Gold to silver sections (approximately 90 nm in thickness) were cut for vessel morphometry. Sections were stained with uranyl acetate and lead citrate and viewed with a Philips 300 TEM).
Morphometry Photographs of suitable vessels (lumen diameter
301 < 10/zm) were taken at various machine magnifications. In our study, vessels of this size range accounted for the majority of all vessels (greater than 90°7o) seen in the tumor tissue. This agrees closely with values obtained from studies of experimental glioma in rat [12]. Morphometric measurements of wall thickness, mitochondrial density and vesicle density were calculated as previously described [13, 14]. Wall thickness is half the difference between the calculated lumen and profile diameters, and includes the endothelial cell, basal lamina, and pericytes (if present). Mitochondrial density is the percent area of endothelial cytoplasm, exclusive of nucleus, occupied by mitochondria. Vesicle density is obtained by counting the number of vesicles present in a test area of endothelium (machine magnification 33 000 x ) expressed as the mean number in one ~m 2 of endothelial cytoplasm. Photographs of transversely sectioned junctions were taken at machine magnification 33000x. They were examined for portions composed of wide segments or clefts, where the adjacent cell membranes are separated by a space of at least 10 nm, versus tight regions, where the outer leaflets of adjacent cell membranes are fused (occluding junctions). It is thought that junctional clefts may represent sections through paracellular channels that wind between occluding junctions [15-18] thus providing a route for extravasation of large solutes. It is essential to appreciate the three-dimensional arrangement of endothelial vesicles in order to understand their role in vascular permeability [13, 19, 20]. Therefore, serial sections of considerable thinness (less than 25 nm) were also cut for control brain and two tumors (patients #1 and #2). Control studies with several tissues from experimental animals showed that vesicle arrangement for vessels within a particular tissue was consistant between animals [13], so values from tumor vessels in the present study were pooled. Consecutive endothelial segments from ten control vessel profiles and 20 tumor vessel profiles were photographed at a machine magnification of 33 000 x . Cytoplasmic organelles were then traced through adjacent sections, and the organization of endothelial vesicles was reconstructed. According to criteria previously described [13], vesicles were clas-
sifted into several types, 'clustered' (attached to at least one other vesicle); 'surface connected' (connected to the endothelial surface by either a short neck or via one or more fused vesicles); 'tubule connected' (fused to golgi, ER or other membranous organelles) and 'free' (no discernable connection to any other organelle or surface). Mean densities of each vesicle class were calculated as the product of the mean vesicle density and the mean relative proportion. Occasionally vesicles were classified into more than one category (i.e. both surface connected and fused to another vesicle). We hypothesize that large clusters of vesicles connected to the endothelial surface represent potential transendothelial channels for extravasation of macromolecules, while numerous free vesicles in the cytoplasm likely indicate that vesicle pinocytosis and shuttling across the cytoplasm accounts for some vessel permeability to large solutes.
Statistical analys& Analysis of variance and Chi square were used to compare mean values among and within groups. Where significant 'F'-ratios were obtained, a modified Tukey's test (T' method) was used to compare means between groups.
Results
For this study, ten or more vessel profiles from at least two blocks of tissue were selected for each tumor, and five vessels were selected from two blocks for each control brain (five vessels per patient). A total of 68 tumor vessels, at least ten from each patient, was analysed.
General morphometry Microvessels selected for this study (Fig. 1) had an average diameter of about six tzm. Also present in tumor tissue, but not sampled were infrequent (less than 10070) very large vessels. Closely associated perivascular cells, probably pericytes but occasion-
302
Fig. 1. Typical microvessels from controlbrain and glioblastoma multiforme a. A microvessel from control brain. Note the thin, continuous endothelium, b. Microvessel from glioblastoma multiforme. Although these tumors have permeable vasculature, their endothelium lacks fenestrations. L = vessel lumen; N = endothelial nucleus; Scale bar = 200 nm.
ally tumor cells, were found associated with 70°7o of the t u m o r vessels, and pericytes were found associated with 80% of the normal barrier vessels. Pericytes have been implicated in the function of the bloodbrain barrier, as a population of potentially phago-
cytic cells [21]. Whether they play a role in these t u m o r vessels is unknown. Mitochondrial density was lowest in control brain vessels and significantly higher in endothelium from t u m o r vessels of patients #2 and #3, and not signifi-
303 Table 1. Morphometric measurements from control brain and tumor microvessels Tissue sample
Sex
Age
Control brain M / F 77/60 Patient 1 M 77 Patient 2 F 43 Patient 3 M 22 Patient 4 F 60 Patient 5 M 77 Values are mean (standard deviation)
Na
Mitochondrial b thickness (um)
Wall thickness ~m)
Vesicle density (//~m2 cytoplasm)
10 10 10 27 11 10
1.5 2.4 4.9 3.8 3.1 3.4
1.44 2.76 1.52 1.06 1.94 1.77
3.7 6.8 5.4 4.8 6.7 2.9
(0.6) (1.7) (0.9) d (1.6) c (1.6) (1.2)
(0.47) (0.89) e (0.55) (0.40) (1.23) (0.43)
(1.6) (3.6) (4.4) (3.5) (3.3) (1.8)
a N u m b e r of vessels examined from each tissue; b As percent of cytoplasm area occupied by mitochondria; c Significantly higher than density from all tissues but patient 2 (p