Neuropaihology and Applied Neurobiology I99 I , 17,177-1 87
Spindle-cell glioblastoma or gliosarcoma? H. J O N E S , P. V. S T E A R T A N D R. 0. W E L L E R Department of Neuropathology, Southampton University Medical School, Southampton General Hospital, Southampton
JONESH., STEART P. V. & WELLER R. 0. (1991) Neuropathology and Applied Neurobiology 17, 177-187 Spindle-cell glioblastoma or gliosarcoma? ‘Gliosarcomas’ have long been considered to be mixed gliomas and sarcomas. The present study failed to define criteria which clearly delineate ‘gliosarcomas’ from glioblastoma multiforme and suggests that ‘gliosarcomas’ should be considered as spindle cell glioblastomas. A total of six cases originally diagnosed as ‘gliosarcomas’ were compared with four cases of glioblastoma multiforme. No clinical or prognostic features were defined which would clearly separate ‘gliosarcomas’ from glioblastoma multiforme. Macroscopically, biopsies from ‘gliosarcomas’ ranged from firm,apparently well-circumscribed tumours to poorly circumscribed lesions with a soft consistency resembling glioblastoma multiforme. Histology revealed a continuous spectrum in which ‘gliosarcomas’ with large reticulin-rich areas of spindle cells merged with typical glioblastomas containing only small islands of spindle cells and reticulin staining. Immunocytochemistry for glial fibrillary acidic protein (GFAP), S 100 protein and a-smooth muscle actin (ASMA) showed that the majority of cells in reticulin-poor areas of ‘gliosarcorna’ and glioblastomas expressed SlOO protein and GFAP; many expressed ASMA and some expressed both GFAP and ASMA. Spindle cells in reticulin-rich areas of ‘gliosarcomas’ and glioblastomas most frequently expressed ASMA but many cells also expressed S100 protein and GFAP; some cells expressed both GFAP and ASMA. The results of this study and a review of the literature suggests that there is a clinical, radiological and pathological continuum with glioblastoma and ‘gliosarcoma’ at different ends of the spectrum. It is suggested, therefore, that most, if not all, ‘gliosarcomas’ be redesignated as spindle cell gIiobIastomas and not be considered as a mixture of glioma and sarcoma. Keywords: gliosarcoma, spindle cell glioblastoma, glial fibrillary acid protein, S 100 protein, alpha smooth muscle actin, immunocytochemistry
INTRODUCTION Stroebe (1895) first suggested that sarcomatous change occurred in glioblastomas and introduced the term ‘gliosarcoma’. Russell and Rubinstein (1989) described these tumours as tough, well-circumscribed lobulated lesions which are often firmly attached to the dura. Microscopically, ‘gliosarcomas’ contain spindle cell areas rich in reticulin together with areas devoid Correspondence to: Professor R. 0.Weller, Department of Pathology, Level E, South Pathology Block, Southampton General Hospital, Tremona Road, Southampton SO9 4XY.
178
H . Jones, P. V . Steart and R. 0. Weller
of reticulin in which the tumour cells have a glial morphology (Morantz,-Feigin & Ransohoff, 1976; Russell & Rubinstein, 1989). However, despite the macroscopic appearances, ‘gliosarcomas’ are usually not well-circumscribed and their prognosis is similar to that of glioblastoma multiforme (Morantz et al., 1976). Despite the distinctive appearance of some ‘gliosarcomas’, there is no firm definition of these tumours and, more particularly, there are no firm criteria by which ‘gliosarcomas’ can be separated from glioblastoma multiforme. It has been suggested that the spindle cell or ‘sarcomatous’ element of gliosarcoma may be a result of exuberant vascular endothelial or fibroblastic proliferation induced by the malignant glial component (Russell & Rubinstein, 1989). However, immunocytochemistry for Factor VIII related antigen has produced conflicting results (McComb et af.,1982; Slowik et al., 1985; Kochi & Budka, 1987) and neither an endothelial nor fibroblastic origin for the spindle cell element in these tumours has been firmly established. The object of the present paper is to determine whether clinical, radiological and pathological criteria can be defined which clearly separate ‘gliosarcomas’ from glioblastoma multiforme or whether most, if not all, ‘gliosarcomas’ should be considered as spindle cell variants of glio blastoma multiforme. Immunocytochemistry in this study includes the use of antibodies against three proteins. The first two proteins, SlOO protein and glial fibrillary acidic protein (GFAP) are characteristically present in glial cells whereas the third protein, a-smooth muscle actin (ASMA) is typically present in vascular smooth muscle and some other mesenchymal cells (Skalli et al., 1986; Jones, Steart & du Boulay, 1990). SlOO protein is an acidic dimeric calcium-binding protein composed of different combinations of alpha and beta subunits and it was first isolated from the central nervous system. It is present in the nucleus and cytoplasm of glial cells, Schwann cells, melanocytes, chondrocytes, adipocytes, myoepithelial cells and in the tumours derived from them (Nakajima et al., 1982: Perentes & Rubinstein, 1987). GFAP, on the other hand, is the protein constituent of intermediate filaments typically present in glial cells of astrocytic origin and in many types of tumour derived from neuroectoderm (Clark, 1984; Russell & Rubinstein, 1989). The antibody to ASMA used in this study was raised by Skalli et al. (1986) to a synthetically produced decapeptide having the sequence of the NH? terminal of alpha smooth muscle actin. Actin is an abundant intracellular protein that polymerizes to form filaments which are essential for cell motility: it is expressed in mammalian cells as six isoforms characterized by amino acid sequence analyses (Garrels & Gibson, 1976). The relative distribution of actin isoforms in smooth muscle differs between organs and different stages of development, but two specific isoforms predominate, namely a-SMA and y-SMA. MATERIALS AND METHODS Ten tumours were included in this study and were selected on the basis of the original pathology report of the surgical biopsy. Six were cases classified as gliosarcoma and four as glioblastoma multiforme. The clinical records of the patients were examined and the CT scans and the reports of the CT scans were reviewed. All tumours were removed surgically through a craniotomy in the Wessex Regional Neurological Centre, Southampton and fixed in 10% neutral buffered formalin for at least 24 hours prior to processing and embedding in paraffin wax. Sections, 5 pm thick, were cut from each case and stained by haematoxylin and eosin, and the Gordon and Sweet technique for reticulin.
Spindle-cell glioblastoma or gliosarcoma?
179
Irnmunocytochernical techniques
Paraffin sections were mounted on glass slides and, foIlowing removal of wax, endogenous peroxidases were inhibited by treating the sections with 0.5% hydrogen peroxide in methanol and this was followed by 8 min incubation in 0. I % trypsin in 0.1 YOCaCI, at 37°C. Sections were then incubated either with a polyclonal antibody to SlOO protein (Dako, High Wycombe, UK) at a dilution of 1:1500, with a polyclonal antibody to GFAP (prepared from GFAP extracted from a cerebellar astrocytoma by the technique of Palfreyman et al. 119791) at a dilution of 1:2000 or with a monoclonal antibody to a-smooth muscle actin (ASMA) (Sigma Chemical Company, Poole, UK) at 1 :4000dilution. All incubations were for 30 min at room temperature. Immunoreactivity was demonstrated by the avidin-biotin peroxidase complex technique (Dako, High Wycombe, UK) and sections were lightly counterstained with haematoxylin. Histological correlations
By the use of a Leitz drawing tube attached to a microscope, reticulin-rich and reticulin-poor areas within serial sections of each tumours 1-3 (Tables 1 and 2) were matched with corresponding areas stained by antibodies to SlOO protein, G F A P and ASMA. Staining in immunocytochemical and reticulin preparations was quantified on a scale to 3 as specified by examples illustrated in Figures 1-3.
+
+
RESULTS Clinical and radiological features
The clinical and radiological features of all 10 tumours are summarized in Table 1. The mean age of presentation of the ‘gliosarcomas’ was 61 years (range 52-72 years). Four patients were male and two were female. This compared with the mean age of 61 years for the presentation of the randomly selected glioblastoma patients, all of whom were female. The length of pre-operative symptoms and post-operative survival are shown in Table 1. CT scans of patients in the ‘gliosarcomas’ group showed a spectrum of features which ranged from discrete well-circumscribed enhancing lesions, indistinguishable from metastases or meningiomas, to poorly defined, patchily enhancing lesions more typical of glioblastoma multiforme. Pathology
Macroscopic examination of the tumours removed surgically from the brain (Table 1) revealed firm areas of tissue in four of the six cases of ‘gliosarcoma’: the consistency was not recorded in one case. Firm consistency was also noted in one of the tumours that was eventually diagnosed as glioblastoma multiforme (Table 1). The other three glioblastomas were soft. Histology and immunocytochemistry
Sections of the six ‘gliosarcomas’ stained with haematoxylin and eosin revealed sheets of spindle cells and islands of tumour cells with a more typically stellate glial morphology. Necrosis was seen in all six tumours. With reticulin staining (Figure la) the spindle cell regions showed an
Age
Sex
Presenration
M
72
63
62
4
5
6
64
62 56
63
7
8 9
10
Glioblastonras
M
52
3
3/12 left side weakness
F
F
2/12 headache. 3/52 memory loss 3/12 fits and hemiparesis
3/52 falling to left, sensory deficit
3/12 loss of balance and headache. 3/7 vomiting 2/52 visual defect, headache, arm weakness 2/12 temporal lobe epilepsy, slurred speech 2/12 speech defect, sleepy
4/52 headache, vomiting and left sided weakness 1/52 right sided weakness?stroke
F
F
M
M
F
62
2
F
54
I
'Gliosarcomas' (spindle cell glioblastomas)
Case
Right parietal SOL, glioma, ring enhancing Left frontal SOL, cystic glioma Right fronto-temporal-low attenuation, haematoma Right posterior frontal cystic glioma
Right frontal SOL,?1"/2", thick-walled ring enhancing Left parasaggital SOL, ? glioma ? meningioma uniform density with irregular enhancement Right parieto-occipital ring enhancing, solid/multilocular, glioma Right occipital SOL, enhancing, irregular glioma Right temporal SOL, low density neoplasm, irregular enhancement Left fronto-temporal?Io/2"
CTscan
Soft white/brown
Grey necrotic soft Brown soft
Firm with necrotic areas
Irregular grey/white firm tumour
Firm tumour greylwhite
Yellow varigated circumscribed
Rubbery varigated firm, nodular
Yellow/white bossilated solid
Not stated
Macro
Table 1. Summary of clinical, radiological and macroscopic features of spindle cell glioblastomas ('gliosarcomas') and glioblastomas
Died 9112 Died 3/12 Died 5/12 Died 12/12
+++ +++
+++ +++
-
-
+/-
Alive 12/12
Alive 18/12 Alive 81 I2
Died 1/12
Died 9/12
+++
++
Died 3/12
Outconie
+++
Rericulin conlenl
P
P
Y
P
0
00
e
Age
Sex
6
62 56 63
4
F F F F
+++ +++ +++ +++
+++
++ +++ +++
+++ ++ +++
+++ +++ +++
++ + ++
++ ++ +++ ++ + ++
na, not applicable -no serial sections on these cases
'No reticulin rich areas present
'Serial sections allowed identification of the same cells
9 10
7 8
Glioblasromas
+++ +++ +++
+++ +++
na na na na
na na na
na na na
na na na na
+ + +
+ + +
Bo I I1 Neither ASMA+ GFAP' ASMA or GFAP'
Serialsections
-
+ +
++
+-
+ + ++ ++ + + + + ++ + + ++ ++ ++ -
na na -
+
-
2
na na -2
na na na
++ ++
Both Neither ASMA + GFAP' ASMA or GFAP'
Serialseclions
+++ + ++ + +++ + +++ na + na + na
ASMA
GFAP SlOO
ASMA
GFAP
SlW
Reticulin rich areas
Reliculin poor areas
'Gliosarcomas ' (spindle cell glioblas t omas) I 54 F 2 62 . F 3 52 M ++ 4 72 M 5 63 M 6 62 M
Case
Table 2. Immunocytochemical staining of 'gliosarcomas' (spindle cell glioblastomas) and glioblastomas for S 100, GFAP and ASMA
182
H . Jones, P. V . Steart and R. 0.Weller
Spindlecell glioblastoma or gliosarcoma?
1 83
intimate network of reticulin (reticulin-rich areas), whereas other areas were almost free of reticulin fibres (reticulin-poor areas). For the most part, glioblastomas showed little reticulin staining except around blood vessels and in areas of capillary endothelial proliferation. However, in case 7, there were small areas in which the tumour cells were spindle-shaped and there was a rich reticulin network: these regions resembled the reticulin-rich areas in a ‘gliosarcoma’ as depicted in Figure la. Patterns of immunocytochemical staining for S 100 protein corresponded very closely to those for GFAP in both ‘gliosarcomas’ and glioblastoma multiforme. When serial sections of ‘gliosarcomas’ were stained for reticulin and G F A P (Figure 11, the reticulin-free areas were seen to contain cells of glial morphology which stained strongly for GFAP. Many of the cells in the reticulin-rich areas also expressed GFAP (Figure Ib); most of the GFAP-positive cells were spindle-shaped but some had a more irregular morphology. In one glioblastoma, case 7, the small areas which were rich in reticutin contained GFAP-positive spindle cells. Such areas were similar to those seen in the ‘gliosarcomas’. Immunocytochemistry for ASMA revealed positively-stained cells in both reticulin-rich areas and in reticulin-poor areas in all six ‘gliosarcomas’. Figure 2 shows serial sections of the same reticulin-rich area of a ‘gliosarcoma’ stained for ASMA (Figure 2a) and GFAP (Figure 2b). Some spindle cells and some irregularly shaped cells expressed both ASMA and GFAP; other cells expressed only one of the proteins, while others expressed neither ASMA nor GFAP. In the reticulin poor areas, up to 80% of the cells with glial morphology expressed both ASMA and GFAP (Figure 3). Sections of all four glioblastomas contained tumour cells expressing both ASMA and GFAP. In some areas, however, cells with glial morphology expressed neither protein. A quantitative estimate of GFAP, SlOO and ASMA expression in all 10 tumours is depicted in Table 2 on a scale + to 3 + in which + = 30% of cells stained, 2 + = 30-60% and 3 + =more than 60% of cells stained in a particular tumour. The grade of staining for GFAP in the reticulin-poor areas in Figure 1 was estimated as 3 whereas the staining in the reticulin-rich area is 2 + . A 3 + level of staining for ASMA was seen in Figure 2a whereas the level of GFAP expression in this reticulin-rich area (Figure 2b) is 2 + . In Table 2, an estimate is given of the proportion of cells expressing the different proteins. In all 10 cases, there were tumour cells which did not express SlOO protein, GFAP o r ASMA. Reactive astrocytes within brain tissue adjacent to ‘gliosarcomas’ expressed S 100 protein and GFAP but did not express ASMA. Hyperplastic blood vessels within ‘gliosarcomas’ and glioblastoma multiforme showed atypical endothelial cells which expressed ASMA. However, in normal blood vessels situated in tissue adjacent to the tumours, there was no staining of the endothelium with anti-ASMA antibody: in these vessels ASMA expression was confined to pericytes and to smooth muscle cells in the media of arteries.
+,
Figure 1. a, Case 2. Spindle-cell glioblastoma (‘gliosarcoma’) showing a reticulin-rich area of tumour (below) and lobulated areas of tumour deficient in reticulin (above). Gordon and Sweet reticulin technique. x 100. b, Same field as a showing positive staining for GFAP in the reticulin-poor regions (above) and many GFAP-rich spindle cells in the reticulin-rich region (below). Blood vessels (bv) show no GFAP staining. Immunocytochemistry for glial fibrillary acidic protein (GFAP). x 100. Figure 2. a, Case 2. Spindle-cell glioblastoma (‘gliosarcoma’). Cells numbered I, 2 and 3 in a reticulin-rich area express ASMA. Cells 1 and 2 also express GFAP (see b); cell 4 is negative for ASMA but expresses GFAP (see b). Mitosis (m). Immunocytochemistry for a-smooth muscle actin (ASMA). x 400.b, Same field as a stained for GFAP. Cells I and 2 express both ASMA and GFAP whereas cell 3 is negative for GFAP. Cell 4 expresses GFAP but not ASMA. Mitosis (m). Immunocytochemistry for GFAP. x 400.
184
H. Jones, P . V. Steart and R. 0 . Weller
Figure 3. a, Case 3. Spindle-cellglioblastoma (‘gliosarcoma’)stained for ASMA. Many of the cells show positive staining and cells 1 and 2 also express GFAP (see b). A pleomorphic glial tumour cell labelled three expresses ASMA but not GFAP. Mitosis (m).Immunocytochemistryfor ASMA. x 400. b, Same field as a stained for GFAP. Cells 1 and 2 express both GFAP and ASMA whereas cell 3 is negative for GFAP. Mitosis (m).Immunocytochemistryfor GFAP. x400.
DISCUSSION The object of the present study was to determine whether there were criteria that would clearly delineate ‘gliosarcoma’from glioblastoma multiforme. The results suggest, however, that there is a clinical, radiological and pathological continuum with spindle cell glioblastomas at one end of the spectrum and the more classical glioblastoma at the other end. It is probably due to this continuous spectrum that clearly defined criteria for ‘gliosarcoma’ have not been determined. Immunocytochemistry in the present study showed that tumour cells in reticulin-rich, spindle cell (so-called sarcomatous) areas of these tumours express the glial proteins S 100 and GFAP together with a-smooth muscle actin (ASMA); similar expression is seen in the more typically glial areas of the tumours. Cells in glioblastoma multiforme also express SlOO protein, GFAP and ASMA. Previous authors (Morantz et af., 1976) have failed to show significant differences in the clinical presentation and survival between patients with spindle cell glioblastoma (‘gliosarcoma’) and the more classical glioblastoma multiforme. Furthermore, in common with our series, a range of CT scan appearances has been recorded by Russell and Rubinstein (1989); although some spindle cell glioblastomas resemble meningioma or metastatic carcinoma on their CT scan images, others resemble classical glioblastoma multiforme.
Spindle-cell glioblastoma or gliosarcoma?
185
The authors of previous histological, electron microscopic and immunocytochemical studies of spindle cell glioblastomas (‘gliosarcomas’) have concluded that the spindle cell element in the reticulin-rich areas of this tumour are derived from endothelial cells or from the cells associated with the adventitia of blood vessels (Feigin & Gross, 1955; Feigin et al., 1958; Rubinstein, 1964, 1972;Peiia & Felter, 1973; Kochi & Budka 1987). However, there is disagreementiregarding the presence of Factor VIII-related antigen. McComb et al. (1982), Kochi and Budka (1987) and Grant et al. (1989) suggest that the spindle cells in the tumour do not express Factor VIII-related antigen whereas Schifferet al. (1984) and Slowik et al. (1985) have reported that this endothelial cell protein is expressed in some areas. Affinity for the ligand of Ulex europeus in the spindle cells is also not conclusive; Slowik et al. (1985) demonstrated affinity whereas Grant et af.(1989) showed no affinity. Alpha-1-antitrypsin has been identified in cells of spindle cell glioblastomas (‘gliosarcomas’) but this may be a reflection of the macrophage population seen within these tumours rather than a histiocytic origin for the tumour cells (Grant et al., 1989). Electron microscopy has failed to demonstrate a conclusive vascular origin for the spindle cell element (Russell & Rubinstein, 1989). It has been reported previously that cells in the spindle cell, reticulin-rich areas express GFAP (McComb et al., 1982; Kochi & Budka, 1987; Grant et al., 1989) but such cells have usually been interpreted as glial cells included within ‘sarcomatous’ areas. Although this is possible, it does not accord with the usual pattern of glial cell inclusion in mesenchyme-rich tissue. The entrapment of astrocytes within the granulation tissue and fibrous tissue around abscesses, for example, usually results in rounded cells in which the glial processes are closely wrapped around the cell body (Weller, 1990). Cells expressing GFAP and SlOO protein in the present scries of spindle cell glioblastomas are spindle or irregular in shape and have long processes. Furthermore the proportion of cells expressing SlOO and G F A P approaches 30% ( 2 + ) in some reticulin-rich areas (Table 2). Although this only accounts for a proportion of the spindle cells, it is well-recorded that many tumour cells in classical glioblastoma multiforme do not express GFAP (Duffy et al., 1980). In an attempt to identify mesenchymal elements within spindle cell glioblastomas (‘gliosarcomas’), an antibody to ASMA was used in this study. ASMA is expressed by some mesenchymal cells and by soft tissue tumours (Jones et al., 1990). In the present series of spindle cell glioblastomas and classical glioblastoma multiforme, spindle cells in reticulin-rich areas, and cells in the more classically glial areas showed variable expression of ASMA, SiOO protein, and GFAP. Some cells expressed all three proteins, whereas others expressed one protein or none of the three proteins. This finding, in particular, suggests that tumour cells in the reticulin-rich and the reticulin-poor areas have a common glial origin. If the spindle cells in gliosarcomas are glial in origin as suggested by the results of the present study, is there evidence that normal and neoplastic glial cells are capable of producing connective tissue elements such as type IV collagen and laminin? Basement membrane is formed by normal and neoplastic astrocytes around blood vessels (Weller, Foy & Cox, 1977; Sapsford, Buontempo & Weller, 1983) and at the surface of the brain as part of the glia limitans (Alcolado et al., 1988). Similarly, type IV collagen and laminin production has been recorded in a GFAPpositive tumour cell line (Alitalo et al., 1983).Astrocytomas rich in reticulin have been described (Russell & Rubinstein, 1989) and classically, reticulin surrounds cells expressing GFAP in pleomorphic xanthoastrocytomas (Kepes, Rubinstein & Eng, 1979) and in desmoplastic cerebral astrocytomas of infancy (Taratuto et al., 1984). Such findings suggest that normal glial cells have the capacity to produce reticulin, and that neoplastic glia also produce reticulin not only in spindle cell glioblastomas (‘gliosarcomas’) but also in other astrocytic tumours.
186
H . Jones, P . V . Steart and R . 0 . Weller
Although endothelial cells in normal brain and blood vessels do not express ASMA, endothelial cells in spindle cell and classical glioblastomas did express this protein that is only usually found in smooth muscle cells of blood vessels. This abnormal expression of a protein may reflect the altered functional role of endothelial cells in glioblastoma blood vessels associated with the alterations in the blood-brain bamer. Alteration in protein expression is also seen in the endothelial cells of blood vessels in oligodendrogliomas which express carbonic anhydrase C whereas the endothelial cells of blood vessels in non-neoplastic brain do not express this enzyme (Weller, Steart & Moore, 1986). Although there are glial tumours which may mimic metastases and meningiomas on CT scan, are firm and appear to be well-circumscribed at surgical excision, the present study suggests that these tumours are spindle cell glioblastomas rather than a mixed gliomatous and sarcomatous tumour (‘gliosarcoma’). The existence of a mixed glial and sarcomatous tumour cannot obviously be excluded in a small series such as ours. But, we have shown here a spectrum of glial tumours ranging from spindle cell glioblastoma to classical glioblastoma multiforme into which tumours in our series fit and most, if not all, the tumours in published reports of ‘gliosarcoma’could be included. We therefore suggest that ‘gliosarcoma’should be considered as a spindle cell variant of glioblastoma multiforme rather than as a mixed glial and sarcomatous tumour. ACKNOWLEDGEMENT The authors express their appreciation to Miss M. Harris for secretarial help in the preparation of the manuscript, REFERENCES Alcolado R., Weller R.O., Parrish E.P. & Garrod D. (1988) T h e cranial arachnoid and pia mater in man: anatomical and ultrastructural observations. Neuropafhologyand Applied Neurobiology 14, 1-17 Alitalo K.. Bornstein P., Vaheri A. & Sage H. (1983) Biosynthesis of an unusual collagen type by human astrocytoma cells in vitro. Journal of Biological Chemistry 258,2656-2663 Clark H.B.(1984) Immunohistochemistry of nervous system antigens. Diagnostic applications in surgical neuropathology. Seminars in Diagnostic Pathology 1,309-316 Duffy P.E., Haung Y.-Y., Rappaport M.M. & Graf L. (1980) Glial fibrillary acidic protein in giant cell tumours of the brain and other gliomas. A possible relationship to malignancy, differentiation and pleomorphism of glia. Acta Neuropathologica (Bert) 52,5 1-57 Feigin I., Allen L.B., Lipkin L. & Gross S.W. (1958) The endothelial hyperplasia of cerebral blood vessels with brain tumors and its sarcomatous transformation. Cancer 11, 264-277 Feigin I.M. & Gross S.W. (1955) Sarcoma arising in glioblastoma of the brain. American Journal of Pathology 31, 633-653 Garrels J.I. & Gibson W. (1976) Identification and characterisation of multiple forms of actin. Ce119,793-805 Grant J.W., Steart P.V., Aguui D., Jones D.B. & Gallagher P.J. (1989) Gliosarcoma: an immunohistochemical study. Acta Neuropathologica (Bert) 79,305-309 Jones H . , Steart P.V. & du Boulay C.E.H. (1990) Alpha smooth muscle actin as a marker for soft tissue tumours: a comparison with desmin. fournal of Pathology 162,29-35 Kepes J.J., Rubinstein L.J.& Eng L.F. (1979) Pleomorphic xanthoastrocytomas: a distinctive meningocerebral glioma of young subjects with relatively favourable prognosis. A study of 12 cases. Cancer 44,1839-1852 Kochi N. & Budka H. (1987) Contribution of histiocytic cells to sarcomatous development of the gliosarcoma. Acta Neuropathologica (Bert) 73,124-1 30 McComb R.D., Jones T.R., Pizzo S.V. & Bigner D.D. (1982) Immunohistochemical detection of F VIII/von Willebrand factor in hyperplastic endothelial cells in gliobastoma multiforme and mixed glioma-sarcoma. Journal of Neuropathology and Experirnenlal Neurology 41,479489
Spindle-cell gfioblastoma or gliosarcoma?
1 87
Morantz R.A., Feigin I. & Ransohoff J. (1976) Clinical and pathological study of 24 cases of gliosarcoma. Journalof Neurosurgery 45,398408 Nakajima T., Watanabe S.. Sat0 Y., Kameya T. & Shimosato Y. (1982) An immunoperoxidase study of SlOO protein in normal and neoplastic tissue. American Journal of Surgical Pathology 6,715-727 Palfreyman J.W., Thomas D.G.T., Ratcliffe J.G. &Graham D.I. (1979) Glial fibrillary acidic protein (GFAP). Purification from human fibrillary astrocytoma, development and validation of a radioimmunoassay for GFAP-like immunoreactivity. Journal of the Neurological Sciences 41, 101-1 13 Peiia E. & Felter R. (1973) Ultrastructure of a composite glioma-sarcoma of the brain. Acfa Neuroparhologica (Berl) 23,90-94 Perentes E. & Rubinstein L.J. (1987) Recent applications of immunoperoxidase histochemistry in human neurooncology. An update. Archives of Pathology Laboratory Medicine 111,796-812 Rubinstein L.J. (1 964) Morphological problems of brain tumors with mixed cell populations. A c f a Neurochirurgica SUPPI. X, 141-165 Rubinstein L.J. (1972) Parhofogy of Tumors of Cenfral Nervous Sysrem. 2nd Series 6th edition. Armed Forces Institute of Pathology. Washington DC. pp. 74-78 Russell D.S. & Rubinstein L.J. (1989) Pathology of Tumours of the Nervous Sysfem. 5th edition. Edward Arnold, London Sapsford I., Buontempo J. & Weller R.O. (1983) Basement membrane surfaces and perivascular compartments in normal human brain and glial tumours. A scanning electron microscope study. Neuropathology and Applied Neurobiology 9, 18 1-1 84 Schiffer D., Giordana M.T., Mauro A. & Migheli A. (1984) GFAP, F VIII/RAg, laminin and fibronectin in gliosarcomas: an immunohistochemical study. Acra Neuroparhologica ( B e d ) 63,108-1 16 Skalli 0.. Ropraz P., Trzeciak A,, Benzonana G., Gillessen D. & Gabbiani G. (1986) A monoclonal antibody against alpha smooth muscle actin: A new probe for smooth muscle differentiation. Journal of Cell Biology 103, 2787-2796 Slowik F., Jellinger K., Gaszo L. & Fischer J. (1985) Gliosarcomas: histological, immunocytochemical, ultrastructural and tissue culture studies. Acfa Neuroparhologica (Berl) 7,201-210 Stroebe H. (1895) Uber Enstehung und Bau der Gehirngliome. Beitrage zur Pathologischen Anafomie und :ur Allgemeinen Pathologie 18,405497 Taratuto A. L.. Monges J., Lylyk P. & Leiguarda R. (1984) Superficial cerebral astrocytoma attached to dura. Report of six cases in infants. Cancer 54,2505-2512 Weller R.O. ( I 990) Bacterial and fungal infections and granulomatous conditions: In Systemic Pathology, Third edition, Vol. 4, Ed. R.O. Weller, pp. 151-179. Churchill Livingstone, Edinburgh Weller R.O., Foy M. & Cox S. (1977) The development and ultrastructure of the microvasculature in malignant gliomas. Neuroparhology and Applied Neurobiology 3,307-322 Weller R.O., Steart P.V. & Moore I.E. (1986) Carbonic anhydrase C as a marker antigen in the diagnosis of choroid plexus papillomas and other tumours: an immunoperoxidase study. In Biology of Brain Tumour, Eds M.D. Walker & D.G.T. Thomas, pp. 115-120. Martinus Nijhoff Publishers, Boston.
Received 20 July 1990 Accepted 13 November 1990
Spindle-cell glioblastoma or gliosarcoma? H. J O N E S , P. V. S T E A R T A N D R. 0. W E L L E R Department of Neuropathology, Southampton University Medical School, Southampton General Hospital, Southampton
JONESH., STEART P. V. & WELLER R. 0. (1991) Neuropathology and Applied Neurobiology 17, 177-187 Spindle-cell glioblastoma or gliosarcoma? ‘Gliosarcomas’ have long been considered to be mixed gliomas and sarcomas. The present study failed to define criteria which clearly delineate ‘gliosarcomas’ from glioblastoma multiforme and suggests that ‘gliosarcomas’ should be considered as spindle cell glioblastomas. A total of six cases originally diagnosed as ‘gliosarcomas’ were compared with four cases of glioblastoma multiforme. No clinical or prognostic features were defined which would clearly separate ‘gliosarcomas’ from glioblastoma multiforme. Macroscopically, biopsies from ‘gliosarcomas’ ranged from firm,apparently well-circumscribed tumours to poorly circumscribed lesions with a soft consistency resembling glioblastoma multiforme. Histology revealed a continuous spectrum in which ‘gliosarcomas’ with large reticulin-rich areas of spindle cells merged with typical glioblastomas containing only small islands of spindle cells and reticulin staining. Immunocytochemistry for glial fibrillary acidic protein (GFAP), S 100 protein and a-smooth muscle actin (ASMA) showed that the majority of cells in reticulin-poor areas of ‘gliosarcorna’ and glioblastomas expressed SlOO protein and GFAP; many expressed ASMA and some expressed both GFAP and ASMA. Spindle cells in reticulin-rich areas of ‘gliosarcomas’ and glioblastomas most frequently expressed ASMA but many cells also expressed S100 protein and GFAP; some cells expressed both GFAP and ASMA. The results of this study and a review of the literature suggests that there is a clinical, radiological and pathological continuum with glioblastoma and ‘gliosarcoma’ at different ends of the spectrum. It is suggested, therefore, that most, if not all, ‘gliosarcomas’ be redesignated as spindle cell gIiobIastomas and not be considered as a mixture of glioma and sarcoma. Keywords: gliosarcoma, spindle cell glioblastoma, glial fibrillary acid protein, S 100 protein, alpha smooth muscle actin, immunocytochemistry
INTRODUCTION Stroebe (1895) first suggested that sarcomatous change occurred in glioblastomas and introduced the term ‘gliosarcoma’. Russell and Rubinstein (1989) described these tumours as tough, well-circumscribed lobulated lesions which are often firmly attached to the dura. Microscopically, ‘gliosarcomas’ contain spindle cell areas rich in reticulin together with areas devoid Correspondence to: Professor R. 0.Weller, Department of Pathology, Level E, South Pathology Block, Southampton General Hospital, Tremona Road, Southampton SO9 4XY.
178
H . Jones, P. V . Steart and R. 0. Weller
of reticulin in which the tumour cells have a glial morphology (Morantz,-Feigin & Ransohoff, 1976; Russell & Rubinstein, 1989). However, despite the macroscopic appearances, ‘gliosarcomas’ are usually not well-circumscribed and their prognosis is similar to that of glioblastoma multiforme (Morantz et al., 1976). Despite the distinctive appearance of some ‘gliosarcomas’, there is no firm definition of these tumours and, more particularly, there are no firm criteria by which ‘gliosarcomas’ can be separated from glioblastoma multiforme. It has been suggested that the spindle cell or ‘sarcomatous’ element of gliosarcoma may be a result of exuberant vascular endothelial or fibroblastic proliferation induced by the malignant glial component (Russell & Rubinstein, 1989). However, immunocytochemistry for Factor VIII related antigen has produced conflicting results (McComb et af.,1982; Slowik et al., 1985; Kochi & Budka, 1987) and neither an endothelial nor fibroblastic origin for the spindle cell element in these tumours has been firmly established. The object of the present paper is to determine whether clinical, radiological and pathological criteria can be defined which clearly separate ‘gliosarcomas’ from glioblastoma multiforme or whether most, if not all, ‘gliosarcomas’ should be considered as spindle cell variants of glio blastoma multiforme. Immunocytochemistry in this study includes the use of antibodies against three proteins. The first two proteins, SlOO protein and glial fibrillary acidic protein (GFAP) are characteristically present in glial cells whereas the third protein, a-smooth muscle actin (ASMA) is typically present in vascular smooth muscle and some other mesenchymal cells (Skalli et al., 1986; Jones, Steart & du Boulay, 1990). SlOO protein is an acidic dimeric calcium-binding protein composed of different combinations of alpha and beta subunits and it was first isolated from the central nervous system. It is present in the nucleus and cytoplasm of glial cells, Schwann cells, melanocytes, chondrocytes, adipocytes, myoepithelial cells and in the tumours derived from them (Nakajima et al., 1982: Perentes & Rubinstein, 1987). GFAP, on the other hand, is the protein constituent of intermediate filaments typically present in glial cells of astrocytic origin and in many types of tumour derived from neuroectoderm (Clark, 1984; Russell & Rubinstein, 1989). The antibody to ASMA used in this study was raised by Skalli et al. (1986) to a synthetically produced decapeptide having the sequence of the NH? terminal of alpha smooth muscle actin. Actin is an abundant intracellular protein that polymerizes to form filaments which are essential for cell motility: it is expressed in mammalian cells as six isoforms characterized by amino acid sequence analyses (Garrels & Gibson, 1976). The relative distribution of actin isoforms in smooth muscle differs between organs and different stages of development, but two specific isoforms predominate, namely a-SMA and y-SMA. MATERIALS AND METHODS Ten tumours were included in this study and were selected on the basis of the original pathology report of the surgical biopsy. Six were cases classified as gliosarcoma and four as glioblastoma multiforme. The clinical records of the patients were examined and the CT scans and the reports of the CT scans were reviewed. All tumours were removed surgically through a craniotomy in the Wessex Regional Neurological Centre, Southampton and fixed in 10% neutral buffered formalin for at least 24 hours prior to processing and embedding in paraffin wax. Sections, 5 pm thick, were cut from each case and stained by haematoxylin and eosin, and the Gordon and Sweet technique for reticulin.
Spindle-cell glioblastoma or gliosarcoma?
179
Irnmunocytochernical techniques
Paraffin sections were mounted on glass slides and, foIlowing removal of wax, endogenous peroxidases were inhibited by treating the sections with 0.5% hydrogen peroxide in methanol and this was followed by 8 min incubation in 0. I % trypsin in 0.1 YOCaCI, at 37°C. Sections were then incubated either with a polyclonal antibody to SlOO protein (Dako, High Wycombe, UK) at a dilution of 1:1500, with a polyclonal antibody to GFAP (prepared from GFAP extracted from a cerebellar astrocytoma by the technique of Palfreyman et al. 119791) at a dilution of 1:2000 or with a monoclonal antibody to a-smooth muscle actin (ASMA) (Sigma Chemical Company, Poole, UK) at 1 :4000dilution. All incubations were for 30 min at room temperature. Immunoreactivity was demonstrated by the avidin-biotin peroxidase complex technique (Dako, High Wycombe, UK) and sections were lightly counterstained with haematoxylin. Histological correlations
By the use of a Leitz drawing tube attached to a microscope, reticulin-rich and reticulin-poor areas within serial sections of each tumours 1-3 (Tables 1 and 2) were matched with corresponding areas stained by antibodies to SlOO protein, G F A P and ASMA. Staining in immunocytochemical and reticulin preparations was quantified on a scale to 3 as specified by examples illustrated in Figures 1-3.
+
+
RESULTS Clinical and radiological features
The clinical and radiological features of all 10 tumours are summarized in Table 1. The mean age of presentation of the ‘gliosarcomas’ was 61 years (range 52-72 years). Four patients were male and two were female. This compared with the mean age of 61 years for the presentation of the randomly selected glioblastoma patients, all of whom were female. The length of pre-operative symptoms and post-operative survival are shown in Table 1. CT scans of patients in the ‘gliosarcomas’ group showed a spectrum of features which ranged from discrete well-circumscribed enhancing lesions, indistinguishable from metastases or meningiomas, to poorly defined, patchily enhancing lesions more typical of glioblastoma multiforme. Pathology
Macroscopic examination of the tumours removed surgically from the brain (Table 1) revealed firm areas of tissue in four of the six cases of ‘gliosarcoma’: the consistency was not recorded in one case. Firm consistency was also noted in one of the tumours that was eventually diagnosed as glioblastoma multiforme (Table 1). The other three glioblastomas were soft. Histology and immunocytochemistry
Sections of the six ‘gliosarcomas’ stained with haematoxylin and eosin revealed sheets of spindle cells and islands of tumour cells with a more typically stellate glial morphology. Necrosis was seen in all six tumours. With reticulin staining (Figure la) the spindle cell regions showed an
Age
Sex
Presenration
M
72
63
62
4
5
6
64
62 56
63
7
8 9
10
Glioblastonras
M
52
3
3/12 left side weakness
F
F
2/12 headache. 3/52 memory loss 3/12 fits and hemiparesis
3/52 falling to left, sensory deficit
3/12 loss of balance and headache. 3/7 vomiting 2/52 visual defect, headache, arm weakness 2/12 temporal lobe epilepsy, slurred speech 2/12 speech defect, sleepy
4/52 headache, vomiting and left sided weakness 1/52 right sided weakness?stroke
F
F
M
M
F
62
2
F
54
I
'Gliosarcomas' (spindle cell glioblastomas)
Case
Right parietal SOL, glioma, ring enhancing Left frontal SOL, cystic glioma Right fronto-temporal-low attenuation, haematoma Right posterior frontal cystic glioma
Right frontal SOL,?1"/2", thick-walled ring enhancing Left parasaggital SOL, ? glioma ? meningioma uniform density with irregular enhancement Right parieto-occipital ring enhancing, solid/multilocular, glioma Right occipital SOL, enhancing, irregular glioma Right temporal SOL, low density neoplasm, irregular enhancement Left fronto-temporal?Io/2"
CTscan
Soft white/brown
Grey necrotic soft Brown soft
Firm with necrotic areas
Irregular grey/white firm tumour
Firm tumour greylwhite
Yellow varigated circumscribed
Rubbery varigated firm, nodular
Yellow/white bossilated solid
Not stated
Macro
Table 1. Summary of clinical, radiological and macroscopic features of spindle cell glioblastomas ('gliosarcomas') and glioblastomas
Died 9112 Died 3/12 Died 5/12 Died 12/12
+++ +++
+++ +++
-
-
+/-
Alive 12/12
Alive 18/12 Alive 81 I2
Died 1/12
Died 9/12
+++
++
Died 3/12
Outconie
+++
Rericulin conlenl
P
P
Y
P
0
00
e
Age
Sex
6
62 56 63
4
F F F F
+++ +++ +++ +++
+++
++ +++ +++
+++ ++ +++
+++ +++ +++
++ + ++
++ ++ +++ ++ + ++
na, not applicable -no serial sections on these cases
'No reticulin rich areas present
'Serial sections allowed identification of the same cells
9 10
7 8
Glioblasromas
+++ +++ +++
+++ +++
na na na na
na na na
na na na
na na na na
+ + +
+ + +
Bo I I1 Neither ASMA+ GFAP' ASMA or GFAP'
Serialsections
-
+ +
++
+-
+ + ++ ++ + + + + ++ + + ++ ++ ++ -
na na -
+
-
2
na na -2
na na na
++ ++
Both Neither ASMA + GFAP' ASMA or GFAP'
Serialseclions
+++ + ++ + +++ + +++ na + na + na
ASMA
GFAP SlOO
ASMA
GFAP
SlW
Reticulin rich areas
Reliculin poor areas
'Gliosarcomas ' (spindle cell glioblas t omas) I 54 F 2 62 . F 3 52 M ++ 4 72 M 5 63 M 6 62 M
Case
Table 2. Immunocytochemical staining of 'gliosarcomas' (spindle cell glioblastomas) and glioblastomas for S 100, GFAP and ASMA
182
H . Jones, P. V . Steart and R. 0.Weller
Spindlecell glioblastoma or gliosarcoma?
1 83
intimate network of reticulin (reticulin-rich areas), whereas other areas were almost free of reticulin fibres (reticulin-poor areas). For the most part, glioblastomas showed little reticulin staining except around blood vessels and in areas of capillary endothelial proliferation. However, in case 7, there were small areas in which the tumour cells were spindle-shaped and there was a rich reticulin network: these regions resembled the reticulin-rich areas in a ‘gliosarcoma’ as depicted in Figure la. Patterns of immunocytochemical staining for S 100 protein corresponded very closely to those for GFAP in both ‘gliosarcomas’ and glioblastoma multiforme. When serial sections of ‘gliosarcomas’ were stained for reticulin and G F A P (Figure 11, the reticulin-free areas were seen to contain cells of glial morphology which stained strongly for GFAP. Many of the cells in the reticulin-rich areas also expressed GFAP (Figure Ib); most of the GFAP-positive cells were spindle-shaped but some had a more irregular morphology. In one glioblastoma, case 7, the small areas which were rich in reticutin contained GFAP-positive spindle cells. Such areas were similar to those seen in the ‘gliosarcomas’. Immunocytochemistry for ASMA revealed positively-stained cells in both reticulin-rich areas and in reticulin-poor areas in all six ‘gliosarcomas’. Figure 2 shows serial sections of the same reticulin-rich area of a ‘gliosarcoma’ stained for ASMA (Figure 2a) and GFAP (Figure 2b). Some spindle cells and some irregularly shaped cells expressed both ASMA and GFAP; other cells expressed only one of the proteins, while others expressed neither ASMA nor GFAP. In the reticulin poor areas, up to 80% of the cells with glial morphology expressed both ASMA and GFAP (Figure 3). Sections of all four glioblastomas contained tumour cells expressing both ASMA and GFAP. In some areas, however, cells with glial morphology expressed neither protein. A quantitative estimate of GFAP, SlOO and ASMA expression in all 10 tumours is depicted in Table 2 on a scale + to 3 + in which + = 30% of cells stained, 2 + = 30-60% and 3 + =more than 60% of cells stained in a particular tumour. The grade of staining for GFAP in the reticulin-poor areas in Figure 1 was estimated as 3 whereas the staining in the reticulin-rich area is 2 + . A 3 + level of staining for ASMA was seen in Figure 2a whereas the level of GFAP expression in this reticulin-rich area (Figure 2b) is 2 + . In Table 2, an estimate is given of the proportion of cells expressing the different proteins. In all 10 cases, there were tumour cells which did not express SlOO protein, GFAP o r ASMA. Reactive astrocytes within brain tissue adjacent to ‘gliosarcomas’ expressed S 100 protein and GFAP but did not express ASMA. Hyperplastic blood vessels within ‘gliosarcomas’ and glioblastoma multiforme showed atypical endothelial cells which expressed ASMA. However, in normal blood vessels situated in tissue adjacent to the tumours, there was no staining of the endothelium with anti-ASMA antibody: in these vessels ASMA expression was confined to pericytes and to smooth muscle cells in the media of arteries.
+,
Figure 1. a, Case 2. Spindle-cell glioblastoma (‘gliosarcoma’) showing a reticulin-rich area of tumour (below) and lobulated areas of tumour deficient in reticulin (above). Gordon and Sweet reticulin technique. x 100. b, Same field as a showing positive staining for GFAP in the reticulin-poor regions (above) and many GFAP-rich spindle cells in the reticulin-rich region (below). Blood vessels (bv) show no GFAP staining. Immunocytochemistry for glial fibrillary acidic protein (GFAP). x 100. Figure 2. a, Case 2. Spindle-cell glioblastoma (‘gliosarcoma’). Cells numbered I, 2 and 3 in a reticulin-rich area express ASMA. Cells 1 and 2 also express GFAP (see b); cell 4 is negative for ASMA but expresses GFAP (see b). Mitosis (m). Immunocytochemistry for a-smooth muscle actin (ASMA). x 400.b, Same field as a stained for GFAP. Cells I and 2 express both ASMA and GFAP whereas cell 3 is negative for GFAP. Cell 4 expresses GFAP but not ASMA. Mitosis (m). Immunocytochemistry for GFAP. x 400.
184
H. Jones, P . V. Steart and R. 0 . Weller
Figure 3. a, Case 3. Spindle-cellglioblastoma (‘gliosarcoma’)stained for ASMA. Many of the cells show positive staining and cells 1 and 2 also express GFAP (see b). A pleomorphic glial tumour cell labelled three expresses ASMA but not GFAP. Mitosis (m).Immunocytochemistryfor ASMA. x 400. b, Same field as a stained for GFAP. Cells 1 and 2 express both GFAP and ASMA whereas cell 3 is negative for GFAP. Mitosis (m).Immunocytochemistryfor GFAP. x400.
DISCUSSION The object of the present study was to determine whether there were criteria that would clearly delineate ‘gliosarcoma’from glioblastoma multiforme. The results suggest, however, that there is a clinical, radiological and pathological continuum with spindle cell glioblastomas at one end of the spectrum and the more classical glioblastoma at the other end. It is probably due to this continuous spectrum that clearly defined criteria for ‘gliosarcoma’ have not been determined. Immunocytochemistry in the present study showed that tumour cells in reticulin-rich, spindle cell (so-called sarcomatous) areas of these tumours express the glial proteins S 100 and GFAP together with a-smooth muscle actin (ASMA); similar expression is seen in the more typically glial areas of the tumours. Cells in glioblastoma multiforme also express SlOO protein, GFAP and ASMA. Previous authors (Morantz et af., 1976) have failed to show significant differences in the clinical presentation and survival between patients with spindle cell glioblastoma (‘gliosarcoma’) and the more classical glioblastoma multiforme. Furthermore, in common with our series, a range of CT scan appearances has been recorded by Russell and Rubinstein (1989); although some spindle cell glioblastomas resemble meningioma or metastatic carcinoma on their CT scan images, others resemble classical glioblastoma multiforme.
Spindle-cell glioblastoma or gliosarcoma?
185
The authors of previous histological, electron microscopic and immunocytochemical studies of spindle cell glioblastomas (‘gliosarcomas’) have concluded that the spindle cell element in the reticulin-rich areas of this tumour are derived from endothelial cells or from the cells associated with the adventitia of blood vessels (Feigin & Gross, 1955; Feigin et al., 1958; Rubinstein, 1964, 1972;Peiia & Felter, 1973; Kochi & Budka 1987). However, there is disagreementiregarding the presence of Factor VIII-related antigen. McComb et al. (1982), Kochi and Budka (1987) and Grant et al. (1989) suggest that the spindle cells in the tumour do not express Factor VIII-related antigen whereas Schifferet al. (1984) and Slowik et al. (1985) have reported that this endothelial cell protein is expressed in some areas. Affinity for the ligand of Ulex europeus in the spindle cells is also not conclusive; Slowik et al. (1985) demonstrated affinity whereas Grant et af.(1989) showed no affinity. Alpha-1-antitrypsin has been identified in cells of spindle cell glioblastomas (‘gliosarcomas’) but this may be a reflection of the macrophage population seen within these tumours rather than a histiocytic origin for the tumour cells (Grant et al., 1989). Electron microscopy has failed to demonstrate a conclusive vascular origin for the spindle cell element (Russell & Rubinstein, 1989). It has been reported previously that cells in the spindle cell, reticulin-rich areas express GFAP (McComb et al., 1982; Kochi & Budka, 1987; Grant et al., 1989) but such cells have usually been interpreted as glial cells included within ‘sarcomatous’ areas. Although this is possible, it does not accord with the usual pattern of glial cell inclusion in mesenchyme-rich tissue. The entrapment of astrocytes within the granulation tissue and fibrous tissue around abscesses, for example, usually results in rounded cells in which the glial processes are closely wrapped around the cell body (Weller, 1990). Cells expressing GFAP and SlOO protein in the present scries of spindle cell glioblastomas are spindle or irregular in shape and have long processes. Furthermore the proportion of cells expressing SlOO and G F A P approaches 30% ( 2 + ) in some reticulin-rich areas (Table 2). Although this only accounts for a proportion of the spindle cells, it is well-recorded that many tumour cells in classical glioblastoma multiforme do not express GFAP (Duffy et al., 1980). In an attempt to identify mesenchymal elements within spindle cell glioblastomas (‘gliosarcomas’), an antibody to ASMA was used in this study. ASMA is expressed by some mesenchymal cells and by soft tissue tumours (Jones et al., 1990). In the present series of spindle cell glioblastomas and classical glioblastoma multiforme, spindle cells in reticulin-rich areas, and cells in the more classically glial areas showed variable expression of ASMA, SiOO protein, and GFAP. Some cells expressed all three proteins, whereas others expressed one protein or none of the three proteins. This finding, in particular, suggests that tumour cells in the reticulin-rich and the reticulin-poor areas have a common glial origin. If the spindle cells in gliosarcomas are glial in origin as suggested by the results of the present study, is there evidence that normal and neoplastic glial cells are capable of producing connective tissue elements such as type IV collagen and laminin? Basement membrane is formed by normal and neoplastic astrocytes around blood vessels (Weller, Foy & Cox, 1977; Sapsford, Buontempo & Weller, 1983) and at the surface of the brain as part of the glia limitans (Alcolado et al., 1988). Similarly, type IV collagen and laminin production has been recorded in a GFAPpositive tumour cell line (Alitalo et al., 1983).Astrocytomas rich in reticulin have been described (Russell & Rubinstein, 1989) and classically, reticulin surrounds cells expressing GFAP in pleomorphic xanthoastrocytomas (Kepes, Rubinstein & Eng, 1979) and in desmoplastic cerebral astrocytomas of infancy (Taratuto et al., 1984). Such findings suggest that normal glial cells have the capacity to produce reticulin, and that neoplastic glia also produce reticulin not only in spindle cell glioblastomas (‘gliosarcomas’) but also in other astrocytic tumours.
186
H . Jones, P . V . Steart and R . 0 . Weller
Although endothelial cells in normal brain and blood vessels do not express ASMA, endothelial cells in spindle cell and classical glioblastomas did express this protein that is only usually found in smooth muscle cells of blood vessels. This abnormal expression of a protein may reflect the altered functional role of endothelial cells in glioblastoma blood vessels associated with the alterations in the blood-brain bamer. Alteration in protein expression is also seen in the endothelial cells of blood vessels in oligodendrogliomas which express carbonic anhydrase C whereas the endothelial cells of blood vessels in non-neoplastic brain do not express this enzyme (Weller, Steart & Moore, 1986). Although there are glial tumours which may mimic metastases and meningiomas on CT scan, are firm and appear to be well-circumscribed at surgical excision, the present study suggests that these tumours are spindle cell glioblastomas rather than a mixed gliomatous and sarcomatous tumour (‘gliosarcoma’). The existence of a mixed glial and sarcomatous tumour cannot obviously be excluded in a small series such as ours. But, we have shown here a spectrum of glial tumours ranging from spindle cell glioblastoma to classical glioblastoma multiforme into which tumours in our series fit and most, if not all, the tumours in published reports of ‘gliosarcoma’could be included. We therefore suggest that ‘gliosarcoma’should be considered as a spindle cell variant of glioblastoma multiforme rather than as a mixed glial and sarcomatous tumour. ACKNOWLEDGEMENT The authors express their appreciation to Miss M. Harris for secretarial help in the preparation of the manuscript, REFERENCES Alcolado R., Weller R.O., Parrish E.P. & Garrod D. (1988) T h e cranial arachnoid and pia mater in man: anatomical and ultrastructural observations. Neuropafhologyand Applied Neurobiology 14, 1-17 Alitalo K.. Bornstein P., Vaheri A. & Sage H. (1983) Biosynthesis of an unusual collagen type by human astrocytoma cells in vitro. Journal of Biological Chemistry 258,2656-2663 Clark H.B.(1984) Immunohistochemistry of nervous system antigens. Diagnostic applications in surgical neuropathology. Seminars in Diagnostic Pathology 1,309-316 Duffy P.E., Haung Y.-Y., Rappaport M.M. & Graf L. (1980) Glial fibrillary acidic protein in giant cell tumours of the brain and other gliomas. A possible relationship to malignancy, differentiation and pleomorphism of glia. Acta Neuropathologica (Bert) 52,5 1-57 Feigin I., Allen L.B., Lipkin L. & Gross S.W. (1958) The endothelial hyperplasia of cerebral blood vessels with brain tumors and its sarcomatous transformation. Cancer 11, 264-277 Feigin I.M. & Gross S.W. (1955) Sarcoma arising in glioblastoma of the brain. American Journal of Pathology 31, 633-653 Garrels J.I. & Gibson W. (1976) Identification and characterisation of multiple forms of actin. Ce119,793-805 Grant J.W., Steart P.V., Aguui D., Jones D.B. & Gallagher P.J. (1989) Gliosarcoma: an immunohistochemical study. Acta Neuropathologica (Bert) 79,305-309 Jones H . , Steart P.V. & du Boulay C.E.H. (1990) Alpha smooth muscle actin as a marker for soft tissue tumours: a comparison with desmin. fournal of Pathology 162,29-35 Kepes J.J., Rubinstein L.J.& Eng L.F. (1979) Pleomorphic xanthoastrocytomas: a distinctive meningocerebral glioma of young subjects with relatively favourable prognosis. A study of 12 cases. Cancer 44,1839-1852 Kochi N. & Budka H. (1987) Contribution of histiocytic cells to sarcomatous development of the gliosarcoma. Acta Neuropathologica (Bert) 73,124-1 30 McComb R.D., Jones T.R., Pizzo S.V. & Bigner D.D. (1982) Immunohistochemical detection of F VIII/von Willebrand factor in hyperplastic endothelial cells in gliobastoma multiforme and mixed glioma-sarcoma. Journal of Neuropathology and Experirnenlal Neurology 41,479489
Spindle-cell gfioblastoma or gliosarcoma?
1 87
Morantz R.A., Feigin I. & Ransohoff J. (1976) Clinical and pathological study of 24 cases of gliosarcoma. Journalof Neurosurgery 45,398408 Nakajima T., Watanabe S.. Sat0 Y., Kameya T. & Shimosato Y. (1982) An immunoperoxidase study of SlOO protein in normal and neoplastic tissue. American Journal of Surgical Pathology 6,715-727 Palfreyman J.W., Thomas D.G.T., Ratcliffe J.G. &Graham D.I. (1979) Glial fibrillary acidic protein (GFAP). Purification from human fibrillary astrocytoma, development and validation of a radioimmunoassay for GFAP-like immunoreactivity. Journal of the Neurological Sciences 41, 101-1 13 Peiia E. & Felter R. (1973) Ultrastructure of a composite glioma-sarcoma of the brain. Acfa Neuroparhologica (Berl) 23,90-94 Perentes E. & Rubinstein L.J. (1987) Recent applications of immunoperoxidase histochemistry in human neurooncology. An update. Archives of Pathology Laboratory Medicine 111,796-812 Rubinstein L.J. (1 964) Morphological problems of brain tumors with mixed cell populations. A c f a Neurochirurgica SUPPI. X, 141-165 Rubinstein L.J. (1972) Parhofogy of Tumors of Cenfral Nervous Sysrem. 2nd Series 6th edition. Armed Forces Institute of Pathology. Washington DC. pp. 74-78 Russell D.S. & Rubinstein L.J. (1989) Pathology of Tumours of the Nervous Sysfem. 5th edition. Edward Arnold, London Sapsford I., Buontempo J. & Weller R.O. (1983) Basement membrane surfaces and perivascular compartments in normal human brain and glial tumours. A scanning electron microscope study. Neuropathology and Applied Neurobiology 9, 18 1-1 84 Schiffer D., Giordana M.T., Mauro A. & Migheli A. (1984) GFAP, F VIII/RAg, laminin and fibronectin in gliosarcomas: an immunohistochemical study. Acra Neuroparhologica ( B e d ) 63,108-1 16 Skalli 0.. Ropraz P., Trzeciak A,, Benzonana G., Gillessen D. & Gabbiani G. (1986) A monoclonal antibody against alpha smooth muscle actin: A new probe for smooth muscle differentiation. Journal of Cell Biology 103, 2787-2796 Slowik F., Jellinger K., Gaszo L. & Fischer J. (1985) Gliosarcomas: histological, immunocytochemical, ultrastructural and tissue culture studies. Acfa Neuroparhologica (Berl) 7,201-210 Stroebe H. (1895) Uber Enstehung und Bau der Gehirngliome. Beitrage zur Pathologischen Anafomie und :ur Allgemeinen Pathologie 18,405497 Taratuto A. L.. Monges J., Lylyk P. & Leiguarda R. (1984) Superficial cerebral astrocytoma attached to dura. Report of six cases in infants. Cancer 54,2505-2512 Weller R.O. ( I 990) Bacterial and fungal infections and granulomatous conditions: In Systemic Pathology, Third edition, Vol. 4, Ed. R.O. Weller, pp. 151-179. Churchill Livingstone, Edinburgh Weller R.O., Foy M. & Cox S. (1977) The development and ultrastructure of the microvasculature in malignant gliomas. Neuroparhology and Applied Neurobiology 3,307-322 Weller R.O., Steart P.V. & Moore I.E. (1986) Carbonic anhydrase C as a marker antigen in the diagnosis of choroid plexus papillomas and other tumours: an immunoperoxidase study. In Biology of Brain Tumour, Eds M.D. Walker & D.G.T. Thomas, pp. 115-120. Martinus Nijhoff Publishers, Boston.
Received 20 July 1990 Accepted 13 November 1990
