Tissue Architecture and Glycosphingolipid Content in Human Gliomas II–IV

Path. Res. Pract. 187, 157-165 (1991)

Tissue Architecture and Glycosphingolipid Content in Human Gliomas II-IV H. D. Mennell, H. Wiegandt2, B. L. Bauer, R. Jennemann, A. F. Rodden 3 and W. Schachenmayr4 1Abteilung Neuropathologie, Zentrum Pathologie, 2Physiologisch-chemisches Institut, 3Abteilung Neurochirurgie, Zentrum fur operative Medizin, Philipps-Universitat, Marburg, FRG; and 4Medizinisches Zentrum fUr Pathologie, Institut fur Neuropathologie, Justus-Liebig Universitat, Giessen, FRG

SUMMARY Contradictory results have been reported claiming either none, partial or almost complete correlation between the complexity of GSL compound profiles and the assumed glial tumor differentiation. Therefore an attempt was made to compare GSL patterns with both the general (final) tumor diagnosis and malignancy grade (WHO) as well as the regional evaluation of the histology and the grading in the tumor tissue pieces directly subjected to biochemical analysis. Regional and general (final) diagnosis did not always correspond, especially when more than one tissue sample of a given tumor was analyzed. Four GSL component patterns were identified by TLC: GSL-type I with gangliosides primarily of the simple Gtac-family lacking sulfatide and the more complex Gtri - and Gtet-gangliosides, GSL-type II with ganglioside of the Gtac- and Gtrdamilies, also without sulfatide, and GSL-type III, with more complex gangliosides of the Gtri - and Gtet-families in addition to Gtac-gangliosides and sulfatide, similar to the normal brain pattern, and the pattern of normal brain. There was only insufficient correlation between these GSL-type patterns and final diagnoses. However, between regional diagnosis of astrocytoma II and GSL-type III on the one hand and glioblastoma multiforme IV and GSL-type I on the other hand, a coincidence of more than 85% was found. In only 50% the intermediate GSL-type II and glioma III were associated. There was no relation between GFAP or vimentin expression and histology or GSL-type both with regard to final and regional diagnoses. Regional astrocytoma architectures exhibiting GSL-type III were mostly fibrillary, whilst glioblastomas with GSL component pattern I had often a giant cell make up. In conclusion regional biochemistry must be compared with the corresponding regional histology and grading, disregarding the overall (general or final) diagnosis, which reflects the most essential or malignant tumor part. Then ganglioside component patterns can be correlated to malignancy grades. This allows the consideration of gangliosides as candidates for malignancy markers in human gliomas.

The abbreviations used for gangliosides follow the systematic notations of Wiegandt21 ; Svennerholm's notation 19 is given in brackets. The glycosphingolipids on the whole were abbreviated as recommended by the International Union of Biochemistry (1976, 1978). © 1991 by Gustav Fischer Verlag, Stuttgart

G stands for ganglioside, the index, e.g. Glac> is the abbreviation for its neutral sugar moiety (lac = lactose, tri = gangliotriaose, tet = gangliotetraose). To this may be added the number of the sialic acid residues, a, b distinguishing between positional sialo isomers. 0344-0338/9110187-0157$3.5010

158 . H. D. Mennel et al. Abbreviations GSL H&E PTAH GF AP AP AAP

= Glycosphingolipids

= Hematoxilin and Eosin (staining)

= Phosphotungstic acid hematoxilin (staining)

= Glial fibrillary acid protein = Alkaline phosphatase anti alkaline phosphatase

(reaction)

CMH

CDH

= ceramide monohexoside = cerami de dihexoside

CTH = cerami de trihexoside CTetH = cerami de tetrahexoside

Introduction Tumor diagnosis may reflect biological (i.e. malignancy) behaviour. Differentiation markers have, however, up to now failed to yield adequate information concerning the degree of malignancy of human intracranial tumors 16 . Glycosphingolipids and in particular gangliosides represent a class of chemically defined compounds that hold promise for use in tumor diagnosis especially in neuroectodermal tissues since these molecules: - are antigenic - are cell surface membrane constituents and since - their structural complexity correlates with the state of differentiation and dedifferentiation. Furthermore positive immunotherapy results involving tumor-associated ganglioside carbohydrate epitopes have been obtained in tumors which were cytogenetically related to intracranial neoplasia 5 . Gtri2 (GD 2) has been identified in neoplastic and reactive astrocytes and consequently in astrocytomas of all grades as well as glioblastoma multiforme 3. Several authors have found an altered profile of glioma gangliosides depending on the grade of malignancy2, 11,22, others have not 8 or have found changes only between benign and highly malignant forms 2o . Since our own results are comparable with those who could establish a correlation between the ganglioside component profiles and the rather benign and the most malignant gliomas, we have correlated the tissue architecture and cellularity to the respective glycosphingolipid patterns in more detail. Particular attention was directed to whether or not the GSL-type established for one particular tumor was the same regardless of the regional tissue and cell make up of the sample analyzed biochemically. To this end, tumors were diagnosed and graded as a whole entity as well as regionally. This approach was chosen, because especially in human gliomas, an increasing regional heterogeneity is observed with advancing malignancy14,16,23; the question therefore arises whether GSL-types are a product of the single variable parts of the tumor or of the neoplasm as a whole.

Material and Methods Tumors were removed from patients and immediately processed. Parts for subsequent biochemical analysis were deep frozen and later treated as follows: Small sections of the tumor

were isolated and divided into two aliquots, which were 5 to 25 mg net weight. In cases where the tissue appeared homogeneous throughout, particles up to 200 mg were used. One half of the material was provided for histology (regional diagnosis), the other for GSL analysis. The remaining bulk of the tumor was analyzed with various conventional methods in order to establish a diagnosis. The following morphological methods were employed: - conventional histology with paraffin embedding and H & E and cresyl violet staining - neurohistological special stainings, e.g. Kanzler's crystal violet stain for glial fibers, Bodians silver impregnations, PTAH and others. - immediate cytological diagnosis from smear preparations stained with methylene blue - frozen sections for instant diagnosis - immunohistochemistry: we made use of a panel of either monoclonal or polyclonal antibodies against differentiation markers or neurogenic tumor markers. These were antibodies against (common) cytokeratin, desmin, vimentin, glial fibrillary acid protein, neurofilament protein; S-100 protein, fibronectin, basic myelin protein and neuron-specific enolase. - electron microscopic analysis. In some tumors, especially those in which the diagnosis seemed clear enough, only some or few of the methods (cytology, histology, special staining, immunohistochemistry, electron microscopy) were performed. Most of them were applied upon separated tumor pieces. Taken together, they all established the final diagnosis and grading of the whole tumor.

Biochemical Methods Methods of GSL analysis were as described before 10. In brief, lyophilised tumor tissue was treated as follows: Nonpolar components were removed by aceton extraction; after centrifugation, the samples were twice sonified in the presence of chloroform-methanol-water, again centrifuged and the solvent extract dried. Saponifiable lipids were hydrolyzed at 500 for five hours at PH 11-12. The separation of neutral and acidic components was performed by ion exchange chromatography following the method of Ledeen 11 • Neutral GSLs were then dried, peracetylated by resuspension in chloroform/acetic anhydride/dimethylaminopyridine, incubated and dried again after addition of toluene. These samples were redissolved in dichloroethane-n-hexane and layered onto small columns filled with Florisil and eluted with dichloroethaneacetone. After drying anew, the samples were transferred to pyrex vials with chloroform/methanol and de-o-acetylated by the addition of sodium methylate in methanol. Salts were then removed by application to Sep Pac RP-18 cartridges and the desalted GSLs eluted with methanol. Gangliosides and purified neutral glycosphingolipids were separated into the different components by thin layer chromatography on HPTLC-Si 60 plates using running solvents of chloroform-methanol-water. The bands were visualized with resorcinolHCL reagent for the acidic and orcinol-sulfuric acid for the neutral components. For additional identification of single glycosphingolipid components on TCL, immunostaining with murine monoclonal antibodies against selective gangliosides was carried out using the APAAP method 10 • Fifteen gliomas of grades II-IV (WHO) were analysed with regard to both their regional histology and their glycosphingolipid content in one to three small tissue portions. There were three astrocytomas of grade II, four astrocytomas, oligodendrogliomas or mixed gliomas of grade III and eight glioblastomas of grade IV. From these tumors, 22 tissue samples of small size were analysed

Glioma histology and GSL . 159 as to their regional histology as well as to their GSL-pattern. Ten of the fifteen tumors, six glioblastomas amongst them, have been investigated in only single samples. Five were investigated more than once, i.e. three tumors in two and two in three samples. Great care was taken to ensure that pieces that had undergone histological examination were exactly the same as the ones used for biochemical analysis. The final diagnoses of the whole tumors were made by either one of the authors (H. D. M. or W. S.), whilst diagnoses of regional histologies were only done by H. D. M.

Results

Morphological Examination In seven out of the fifteen cases the diagnoses of the regional histology and that of the whole tumor were identical, but only one of them had been analyzed in more than one probe. In this instance, two regional histologies were astrocytoma II as the whole tumor (Table 1, No. 3.1-2). In eight of the fifteen, there was complete or partial discrepancy between final and regional histologies. These concerned obviously the remaining four tumors with more than one regional histological analysis (Table 1, Nos. 2.1-3,4.1-2,6.1-2, 10.1-3). Since the most malignant part of the tumor defines its final diagnosis, the regional histology must be less malignant, if deviating, from the final diagnosis. The inverse was the case in one instance only, when the regional histology was thought to represent astrocytoma III, whilst the final diagnosis was astrocytoma II (Table 1, No. 8.1).

GSL Component Distribution Four typical TLC-profiles of GSL-compounds were found in the respective tumors samples: The GSL-pattern of normal brain contained in the neutral compartment almost exclusively the CMH, galactosylceramide. The acidic fraction GSL consisted of the gangliosides Gl ac -, Gtr ;- and Gtet-families as well as sulfatide. The GSL-type III pattern was similar to that of normal brain. However, the neutral fraction GSL contained, in addition to galactosylceramide glucosylceramide as CMH, CDH and CTH and traces of CTetH. All gangliosides and the sulfatide found in normal brain were also present in GSL-type III. Yet, the G1ac- and Gtr;-compounds were increased when compared to the pattern of the normal brain. The GSL-type II pattern showed within the neutral compartment a similar component profile as GSL-type III, except for a considerable reduction of CMH; the gangliosides were of the Gl ac - and Gtr;-family, no Gtet-gangliosides and no sulfa tide being present. The GLS-type I pattern was the simplest of all with regard to the gangliosides. The neutral fraction GSL was similar to that of GSL-type II, yet CMH was almost completely absent. Within the acidic group, the spectrum was restricted almost entirely to the occurence of the Glac-gangliosides Glac1 and Glac2 (GM3 and GD3). Sulfatide was absent, other minor acidic compounds were present in trace amounts. Thus, an increasing simplification from normal brain ganglioside pattern to that of GSL-type III, II to I was observed, comprising:

Table 1. Correspondence of histological grading and GSL-profilec in human gliomas (for GSL-type see text) No. 1.1 2.1 2.2 2.3 3.1 3.2 4.1 4.2 5.1 6.1 6.2 7.1 8.1 9.1

10.1

10.2 10.3 11.1 12.1 13.1 14.1 15.1

Tumor No.

Final Diagnosis

Regional histology and grading

Description

GSL-Type

52/88 69/88 69/88 69/88 77/88 77/88 47/88 47/88 73/88 46/88 46/88 48/88 74/88 45/88 33/88 33/88 33/88 34/88 38/88 50/88 25/88 16/88

Glioblastoma IV Astrocytoma III Astrocytoma III Astrocytoma III Astrocytoma II Astrocytoma II Glioblastoma IV Glioblastoma IV Astrocytoma II Oligodendroglioma III Oligodendroglioma III Oligodendroglioma III Astrocytoma II Glioblastoma IV Glioblastoma IV Glioblastoma IV Glioblastoma IV Glioblastoma IV Glioblastoma IV Mixed glioma III/IV Glioblastoma IV Glioblastoma IV

BRAIN BRAIN ASTRO II ASTRO II ASTRO II ASTRO II ASTRO II ASTRO II ASTRO II ASTRO II ASTRO III ASTRO III ASTRO III GUO IV ASTRO III GLIO IV GUO IV GUO IV GUO IV GUO IV GUO IV GLIO IV

Reactive gliosis Reactive gliosis Fibrillary astrocytoma Fibrillary astrocytoma Fibrillary astrocytoma Fibrillary astrocytoma Fibrillary astrocytoma Fibrillary astrocytoma Fibrillary astrocytoma Fibrillary astrocytoma Small cell astrocytoma Protoplasmic astrocytoma Protoplasmic astrocytoma Giant cell glioblastoma Protoplasmic astrocytoma Small cell glioblastoma Mixed cell glioblastoma Giant cell glioblastoma Giant cell glioblastoma Giant cell glioblastoma Giant cell glioblastoma Giant cell glioblastoma

0 0

III

III III III III III III II III II II II I I I I I

r I I

160 . H. D. Mennel et al.

- the decrease and loss of CMH in the form of galactosylceramide and the increase of CDH, CTH and other "higher" neutral GSL including the CTeTH neolactotetraosylceramide to detectable levels, - the loss of sulfatide, present in normal brain and GSL-type III, but absent in the two other GSL-types - the continuous relative decrease of Gtet and Gtri and the increase of Gl ac gangliosides.

Correlation of GSL-Type to the Final (General) Diagnosis Tumor Group Glioma IV and III-IV: Glioblastoma multiforme (8 cases) in five instances showed GSL-type I, in one case each, GSL-types II, III or the pattern of normal brain. Mixed glioma III1IV (1 case) also was of the GSL-type I.

Figs. 1-4. 1: Fibrillary architecture of astrocytoma II correlated with GSL-type III. Loose network formed by fibrillary astrocytes; between the cells and their processes, numerous microscopical cysts are formed. Regional diagnosis: fibrillary astrocytoma II, general diagnosis: Polymorphic oligodendroglioma. N 46/88, H & E, x 39. - 2: Very loose arrangement of tumor cells with larger empty spaces in between. Regional and general diagnosis: Fibrillary astrocytoma II. GSL-type III. N 77/88, cresylviolet, x 39. - 3: GFAP expression in some neoplastic cell somata and their extensions. Regional diagnosis: Fibrillary astrocytoma II, general diagnosis: Astrocytoma III. GSL-type III. N 69/88, GFAP-PAP, X 39. - 4: Vimentin expression in this case (same as 3) seems equally strong. A few more isolated processes are depicted. Regional diagnosis: Fibrillary as-

trocytoma II, general diagnosis: Astrocytoma III. N 69/88, Vimentin-PAP, X 39.

Glioma histology and GSL . 161

Glioma III: Oligodendroglioma III (two cases) displayed GSL-types II and II and III respectively. One astrocytoma III analyzed in three different portions was characterized by GSL-type III (twice) and normal brain. Glioma II: One astrocytoma II yielded, when analyzed in one and another in two portions, the GSL-type III pattern, whereas a third tumor of this type showed as a single sample the GSL-type II. From these observations it can be concluded that there is a relative preponderance (62.5%) of GSL-type I in glio-

Figs. 5-8. 5: GFAP and vimentin expression in protoplasmiclgemistocytic cells of astrocytoma III and glioblastoma yielding GSL patterns II and I. Protoplasmic astrocytes in partly cystic astrocytoma III (regional) to II (general) with most of the cells stained. N 74/88, GFAPimmunogold staining, x 16. - 6: Weaker reaction with only some stained cells in the same tumor as in Figure 5. N 74/88, GFAP-immunogold staining, x 16. - 7: GFAP reaction in giant cells and gemistocytes of glioblastoma multiforme. Regional and general diagnosis: Glioblastoma multiforme, GSLtype I, N 38/88, GFAP-PAP, x 40. - 8: Vimentin expression in the same tumor as in Figure 7 shows almost identical pattern. N 38/88, Vimentin-PAP, x 40.

blastoma multiforme, if one had to rely only on the final diagnosis and a preponderance of GSL-type III in astrocytoma II (66.6%). Yet, there is broad variation in the remaining glioblastomas and a lack in correlation for the other diagnoses.

Correlation of Regional Histology to GSL-Types The correspondence between histology (and in the same place grading) and GSL-type was much better, when the

162 . H. D. Menne! et al.

regional histological pattern only was compared: Normal

brain, white and gray matter always exhibited their respective normal brain GSL-patterns, regardless of whether the final diagnosis was tumor or not. In both cases, where the brain was under investigation, but the final tumor diagnosis was either anaplastic astrocytoma or glioblastoma, a marked reactive gliosis within the specimen was found. This event, most probably due to the fact that we deal with brain adjacent to tumor (BAT), did not

alter the GSL distribution to any marked extent. This observation is in contrast with other previous reports 20 • In eight cases of astrocytoma II regional diagnosis (Figs. 1-4), GSL-type III was found seven times. The remaining case exhibited type II. This is a 87.5% association of astrocytoma II (WHO) regional diagnosis with the described GSL-type III. Similarly, seven out of eight pieces of glioblastoma multiforme regional architecture (Figs. 13-16) had the most simple arrangement of acidic and

Figs. 9-12. 9: Gemistocytic cytology associated with GSL-type less than III, occurring in astrocytoma III and glioblastoma: Regional and general diagnosis: Glioblastoma multiforme, N 38/88. Rounded gemistocytes with stuffed cytoplasm, H & E, x 40. - 10: Astrocytic processes are visualized by means of glial fibrillary stains: PTAH, x 40. Same case as Fig. 9.11: Electron microscopy reveals abundance of glial filaments within the cytoplasm which give positive reaction to anti-GFAP and antiVimentin (same case as Figs. 9 and 10), x 2000. -12: Higher magnification of Fig. 11. The density of glial filaments is so strong that

intracytoplasmic condensation centers are formed corresponding to Rosenthals fibers, x 20000.

Glioma histology and GSL . 163

neutral GSL, i.e. pattern I. One remaining area diagnosed as glioblastoma had pattern II. There were some difficulties with GSL-pattern II, which could twice be ascribed to regional astrocytic architecture grade III, and in one instance each to grade II and IV. Thus, GSL-type II obviously takes an intermediate position as does glioma grade III; yet, the two did not coincide by more than 50%. In one instance only, astrocytoma grade III showed the more benign GSL-type III, that otherwise was

Figs. 13-16. 13: Giant cell glioblastoma associated with GSL-type I: Giant cells are very conspicuous, but most of the cells are smaller. Regional and general diagnosis: Glioblastoma. N 16/88, cresylviolet, X 16.-14: Another example of giant cell glioblastoma. Regional and general diagnosis: Glioblastoma. GSL-type I. N 25/88, cresylviolet, X 16. - 15: Higher magnification of Fig. 14 shows that giant cells, sometimes multinuclear, taper with their processes into a network formed by many other smaller neoplastic cells. N 25/88, H & E, X 40. - 16: Ultrastructurally, nuclei are often bizarre, the cytoplasm contains abundant filaments, mitochondria and ribosomes. N 25/88, x 7000. Same case as Figs. 14 and 15.

found only in astrocytoma grade II. On the other hand, one specimell'Ofastrocytoma III architecture was associated with a GSL-type I pattern. In addition, for gliomas of different grades an even closer association between tissue architecture and GSL pattern was found. All astrocytomas with GSL-type III had a very similar cellular and tissue make up. They belonged to the micro cellular, loosely arranged fibrillary type, one of the classic forms of relatively benign glioma (Figs. 1,2). In

164 . H. O. Mennel et al.

contrast, GSL-type II in astrocytoma of grade II and III was preferentially associated with formation of protoplasmic and/or gemistocytic forms of tumor cells (Figs. 9-12). All but one glioblastomas with giant cells (Figs. 13-16) had GSL-type I. On the other hand, in two cases, where small globuliform cells and a mixed cell population in glioblastoma were found, equally type I resulted. It is remarkable that large necroses, vessels and/or bleeding did not significantly influence the result. There was no correspondence of GF AP (glial fibrillary acid protein) or vimentin expression to either grade or GSL-type observed. Some of the astrocytomas had weak, some intermediate and some strong reaction to anti-GFAP antibodies (Figs. 1, 2, 5-8). The same held true for vimentin, which was generally present to some extent in astrocytomas. Two giant cell glioblastomas (GSL type I) did not express GFAP or vimentin at all. In the remaining tumors they were expressed in various extents (Figs.

11,12).

Discussion Altered expressions of cellular GSL have long been known to be associated with tumor malignancy; cerebral tumors are no exception. Attempts to correlate ganglioside content and tumor grade have yielded varying results. Yates 22 stated that it was necessary to "examine tissues closely adjacent to those analysed chemically". We have attempted to do this in our study. Several generalisations can be drawn from our results. 1) If one further relies upon biochemical bulk analysis, i.e. analysis of brain or tumor mass without taking into consideration its internal morphological composition, one is bound to use small samples with clearly defined histology following the above mentioned principles. We have shown that the correlation of malignancy grades to GSL patterns, thought to represent steps of (de)differention, increases by adopting the mentioned proceeding. 2) There is now good evidence that the most malignant gliomas (glioblastomas) show a defined GSL pattern 2, 7,20 which allows their identification and separation from more benign tumors. The present study shows that glioblastoma multiforme almost invariably (87.5%) displaya clear pattern distinct from others (GSL-type I). 3) There are additional indications that two other GSL-types also may be related to glioma malignancy grades. The described GSL-type III is highly correlated to fibrillary astrocytoma II. The intermediate GSL-type II could be paralleled with histologic intermediate grade III, but the actual coincidence in this last case was only 50%. 4) There is however, no complete consistency for any type of glioma grade to GSL-type or vice versa. Inconsistencies are especially frequent when comparing the intermediate stages of biochemical results and histological grading, i.e. GSL-type II vs glioma grade III. This is not surprising since intermediate forms always pose the questions of clear cut delineations.

5) Not surprisingly, our results suggest that the regional tissue is better reflected in the GSL-distribution than the final and general tumor diagnosis. Furthermore, there is good correlation between defined tissue architecture and GSL-type: GSL-type I is connected with giant cell glioblastoma, GSL-type III with fibrillary astrocytoma. This may be taken as an indication that the glial cell type is indeed the major factor in determining the respective ganglioside component distribution. This last point deserves some further consideration: It seems in fact tempting to deduce from our results some kind of cellular malignancy scale, which could be fitted to conventional diagnoses: Fibrillary astrocytoma as benign (WHO II - GSL-type III), gemistocytic astrocytoma as intermediate (WHO II-III - GSL-type II) and giant cell glioblastoma as malignant (WHO IV - GSL-type I). This is all the more tempting, since there are definite similarities of gemistocytic cytology to that of more malignant astroblastoma formations and it had repeatedly been claimed that gemistocitic astrocytoma behaves less benign than other forms 6,13 and that they generally develop into glioblastoma 1. Yet, even though not excluding this possibility, our findings are not uniform enough to corroborate such a view. There are additional findings concerning the role of giant cells, which are not in aggreement with the mentioned assumption4 . In conclusion, the results presented here allow the assumption of a positive correlation between increasing glioma malignancy and decreasing ganglioside complexity. In gliomas II-IV a tentative correspondence with distinct GSL-types III-I could also be established. If GSL-content proves valuable as an indicator of malignancy, comparison with other proliferation markers must be attempted. It must be further elucidated whether the correlation GSL-type to malignancy grade could be established in other intracranial tumors, particularly in (pilocytic) astrocytoma of grade I. However, even if this were not the case, the corroboration of three differing patterns of GSL in the major intracranial neoplasia, glioma I-III might expedite better definition of malignancy grades these intracranial tumors. References 1 Ayala AG, Mackay B (1974) Pathology of gliomas. Cancer Bull 26: 82-86 2 Berra B, Gaini SM, Riboni L (1985) Correlation between ganglioside distribution and histological grading of human astrocytomas. Int J Cancer 36: 363-366 3 Bosslet K, Mennel HD, Rodden AF, Bauer BL, Wagner F, Altmannsberger A, Sedlacek HH, Wiegandt H (1989) Monoclonal antibodies against epitopes on ganglioside GD 2 and its lactones. Cancer Immunol Immunother 29: 171-178 4 Burger PC, Vollmer RT (1980) Histologic factors of prognostic significance in the glioblastoma multiforme. Cancer 46: 1179-1186 5 Cheung NKV, Saarinen, UM, Neely JE, Landmeier B, Donovan D, Coccia PF (1985) Monoclonal antibodies to a glycolipid antigen on human neuroblastoma cells. Cancer Res 45: 2642-2649

Glioma histology and GSL . 165 6 Elvigde AR, Martinez-Coli A (1956) Long-term follow-up of 106 cases of astrocytoma 1928-1939. J Neurosurg 13: 230-243 7 Eto Y, Shinoda S (1982) Gangliosides and neutral glycosphingolipids in human brain tumors: Specificity and their significance. Adv Exp Med Bioi 152: 279-290 8 Fredman P, von Holst H, Collins VP, Granholm L, Svennerholm L (1988) Siallylactotetraosylceramid, a ganglioside marker for human malignant gliomas. J Neurochem 50: 912-919 9 GerdesJ, Schwab U, Lemke H, Stein H (1983) Production of mouse monoclonal antibody reactive with a human nuclear antigen associated with cell proliferation. Int J Cancer 31: 13-20 10 Jennemann R, Rodden F, Bauer BL, Mennel HD, Wiegandt H (1990) Glycosphingolipids of human brain tumors I: Gliomas. Cancer Res (in press) 11 Kostic D, Buchheit F (1970) Gangliosides in human brain tumors. Life Sciences 9: 589-596 12 Ledeen RW, Yu KR, Eng LF (1973) Gangliosides of human myelin: sialylgalactsylceramide (G 7) as a major component. J Neurochem 21: 829-839 13 Leibel SA, Sheline GE, War a WM, Boldrey EB, Nielsen SL (1975) The role of radiation therapy in the treatment of astrocytoma. Cancer 35: 1551-1557 14 McComb RD, Bigner DD (1984) The biology of malignant gliomas - a comprehensive survey. Clin Neuropathol 3: 93-106

15 Mennel HD (1988) Geschwulste des zentralen und peripheren Nervensystems. In: Doerr W, Seifert G, Uehlinger F (Hrg) Spezielle pathologische Anatomie. Springer, Berlin-Heidelberg-New York, 13, III, S. 214-542 16 Mennel HD, Bender 0 (1989) Das Glioblastom: Morphologische Beitrage zu einem Problemtumor. Nervenheilk 8: 152-160 17 Nishizaki A, Orita T, Ikemaye Y, Aoki H, Sasari K (1989) Flow - cytometric DNA analysis and immunohistochemical measurement of Ki-67 and BUdR labeling indices in human brain tumors. J Neurosurg 70: 379-384 18 Plate KH, Arndt D, Hellwig D, Ruschoff J, Mennel HD (1989) Assessment of histogenesis and proliferative potential in cytological specimens of human brain tumors: Value of immunocytochemistry and nucleolar organizer regions. Acta Cytol (in press) 19 Svennerholm L (1963) Chromatographic separation of human brain gangliosides. J Neurochem 10: 613-623 20 Traylor TD, Hogan EI (1980) Gangliosides of human cerebral astrocytomas. J Neurochem 34: 126-131 21 Wiegandt H (1985) Gangliosides. In: Neuburger A, van Deenen, L. L. M. (Eds.). New Comprehensive Biochemistry. Elsevier, Amsterdam, Vol 10, pp. 199-260 22 Yates AJ (1988) Glycolipids and gliomas - A review. Neurochemical Pathology 8: 157-180 23 Zulch KJ, Mennel HD (1974) The biology of brain tumors. In: Vinken P], Bruyn GW (Eds.) Handbook of Clinical Neurology, Vol 16, North Holland/Elsevier, Amsterdam, pp 1-55

Received March 19, 1990 . Accepted June 12, 1990

Key words: Glioma II-IV (WHO) - Regional histology - Glycosphingolipids - Glial fibrillary acid protein Gangliorides Prof. Dr. H. D. Mennel, Abteilung Neuropathologie, Zentrum fur Pathologie, Klinikum Lahnberge, Philipps-Universitat, BaldingerstrafSe, D-3550 Marburg, FRG