Continuous improvement of neuroimaging techniques has made it possible to detect ever smaller tumors of the central nervous system (CNS) and greatly increased the chance of an early diagnosis. Examination of the cerebrospinal fluid (CSF) for tumor cells or tumor-suspect cells is a much less sensitive method of tumor detection, for reasons explained below. Yet, despite the availability of advanced neuroimaging studies, cytological examination of the CSF still retains a useful role in the diagnostic evaluation of brain tumors. Not all types of brain tumor can be detected and reliably assigned to a particular diagnostic category by neuroimaging alone. The evaluation of cellular changes in the CSF in primary and metastatic brain tumors, particularly those that infiltrate the meninges, in the light of our rapidly expanding knowledge in immunocytology and molecular genetics, not only facilitates diagnosis but, increasingly often, also provides important information for the choice of an appropriate, specific treatment strategy.

Although cytological examination of the CSF is mainly directed toward the detection of tumor cells, or at least of tumor-suspect cells (neoplastic meningitis [or neoplastic meningitis]), it can also provide important additional findings, such as evidence of tumor-related hemorrhage or inflammatory reactions. It is an indispensable method of monitoring the effect of treatment in a number of conditions (leukemia, lymphoma, etc.).

These general remarks will serve as an introduction for the following, more detailed discussion.

The propensity of cells from primary and secondary CNS tumors to exfoliate (slough off) and migrate in the CSF depends on their degree of malignancy (grade), on the local environmental conditions, and on certain limiting factors of the transport pathways. (Some new information on the mechanisms of these processes and on the target sites for CSF colonization can be found in the current literature in neurology and neuropathology.)

The likelihood of finding tumor cells in the CSF rises if the tumor is highly malignant (grade III or higher; rarely, grade II as well), if the local environment favors exfoliation and migration into the CSF, and if the transport of tumor cells into the CSF is not excessively restricted (e.g., cells derived from metastatic tumor in the brain must be able to traverse capillary systems of those organs that separate them from the CSF). Beside this triad of important factors, the spatial proximity of an extra- or intracranial tumor to the CSF space is a another, less important, determinant of whether tumor cells will appear in the CSF.

It is extremely rare for low-grade tumors near the CSF space to deposit tumor cells in the CSF, although such cells are found somewhat more often after surgical resection of the tumor. Activated forms of nonneoplastic cells can appear as a meningeal reaction to tumors of any degree of malignancy. Thus, the CSF cytologist often faces the task of distinguishing true tumor or tumor-suspect cells from non-neoplastic cells. This problem concerns the diagnosis of low-grade tumors in particular. (This will be discussed further below, as will the relative importance of tumor resection in the context of limiting factors of the transport pathways.)

CSF tumor cells, or at least tumor-suspect cells, are often found accompanying tumors of the following types, if their appearance is favored by the triad of factors mentioned above:

Extracranial tumors can metastasize (usually by the hematogenous route, occasionally via the lymphatic vessels) to the leptomeninges, causing neo-plastic (carcinomatous) meningitis, or to the parenchyma of the brain or spinal cord, leading to the formation of parenchymal CNS metastases. On the one hand metastatic disease may be exclusively leptomeningeal, i.e., tumor cells may be found in the CSF although no parenchymal metastases are detectable by imaging studies. The vital importance of cytological examination of the CSF in such cases is obvious. On the other hand, tumor cells may also be found in the CSF in exclusively parenchymal metastatic disease, unaccompanied by neoplastic meningitis. Cells from a parenchymal tumor can exfoliate and migrate through the choroid plexus and the meningeal vessels into the CSF space, as long as the tumor is highly malignant and the local environmental conditions favor these processes.

Leptomeningeal infiltrates of leukemic cells and malignant lymphoma cells often develop into leukemic or lymphomatous meningitis.

Tumors located in the meninges can metastasize along the CSF pathways. Exfoliated cells and cell clusters travel through the CSF circulation to distant sites where they grow into new tumors (“drop metastases”). Meningeal tumors are often accompanied by tumor cells in the CSF.

These general principles will be elaborated in further, more concrete detail below in the sections dealing with individual tumor types. The relatively small number of cases studied to date makes it impossible for us to give reliable percentage estimates of the frequency of various CSF findings for each type of tumor. More global estimates, which are not based on classification by tumor type, but rather apply to larger categories of disease, (such as primary versus metastatic, or benign versus malignant brain tumors), can be found in the literature. However, we consider such information to be of limited practical use.

How do the diagnostic capabilities of the CSF cytologist, working with preparations of CSF stained with the May–Grünwald–Giemsa method, compare with those of the neuropathologist, working with specimens of solid tissue?

The neuropathologist determines the type of tumor that is present and its degree of malignancy by examining a tissue specimen and carefully taking note of certain distinguishing parameters that, broadly speaking, fall into three classes: histopathological (cell density, cellular and nuclear polymorphism, mitotic activity, pathological endothelial proliferation, necrosis of tumor tissue, areas of infiltration), biological (molecular genetic, immunocytochemical, and immunohistochemical markers, etc.), and clinical. These same parameters also underlie the World Health Organization (WHO) classification of the degree of malignancy (anaplasia) of CNS tumors into four grades (I–IV) . Many types of CNS tumor have a clearly definable histopathological architecture and are thus easy to diagnose definitively according to such criteria, yet a considerable number of tumor types remain for which this is not possible. The problem is most obvious for tumors in which cells are poorly differentiated to the extent that their distinguishing features are either atypical or else not specific for any particular tumor type. In such cases, even if immunocytochemical markers (antibodies) are used to detect various tumor cell surface antigens, the neuropathologist will only be able to diagnose an “unclassifiable tumor” or a mixed tumor type.

The CSF cytologist faces much greater difficulties:

Tumor cells are found in the CSF as individual cells or very small cell clusters (only rarely as larger ones). Therefore, the histopathological features of the tumor that are used to define its degree of malignancy and hence its diagnostic classification—that is, the cell density, pathological endothelial proliferation, necrosis of tumor tissue, and areas of infiltration—are not available for inspection. A tumor cell cluster in the CSF, if present, may be an intact clump of tumor tissue that has separated from a tumor lying near the CSF space; alternatively, it may have been formed by the secondary aggregation of tumor cells that were originally present in the CSF as individual cells and then came together under the influence of local environmental conditions promoting cell adhesion.

Tumor cells in the CSF are also subject to a considerable degree of secondary change resulting from the marked difference between their original surroundings in the tissue and their new, liquid environment. Although the cells may still be recognizable as tumor or tumor-suspect cells in a May–Grünwald–Giemsa preparation, these changes may render them diagnostically unclassifiable or permit, at most, an assignment to a broad diagnostic category, rather than a specific type of tumor. The distinction between true tumor cells and nonneoplastic cell types of similar appearance in May–Grünwald–Giemsa preparations, such as irritative forms of other cell types, incompletely differentiated precursor cells of other cell populations, may also be difficult and necessitate the use of specific marker tests. This problem will be discussed further below.

What combination of criteria can the CSF cytologist still use to diagnose a tumor cell, or at least a tumor-suspect cell, in a May–Grünwald–Giemsa preparation?

An important feature of the current WHO classification, as well as of its predecessor (1993), is that they both dispense with the terms “primary brain tumor” and “secondary brain tumor,” even though these terms were in use for many years and are still familiar to CSF cytologists from textbooks of neurology.

For the sake of completeness, it should be mentioned that the literature contains reports of tumor cells in the CSF in certain kinds of tumor for which we have, to date, found no tumor cells at all or else, in very rare cases, cells with only a remote suspicion of neo- plasia in the CSF. In our experience, further study has usually revealed cells of the latter type to be atypical and unclassifiable, rather than neoplastic. We refer the reader to the relevant literature for each of the tumor types for which this is true. Tumor-suspect cells have been reported in a few cases of meningioma (Dufresne 1972), anaplastic oligodendroglioma (Watson and Hajdu 1977), and neuroblastoma (Gandolfi 1980). No CSF cells with malignant features, but, at most, atypical forms have been reported in craniopharyngioma and neurinoma.

In view of the limited diagnostic potential of the May–Grünwald–Giemsa method, it is appropriate to ask under what circumstances immunocytochemical tumor marker tests are to be considered necessary, advisable,or superfluous for the purposes of clinical diagnosis and treatment.

Such tests are necessary when a routine May–Grünwald–Giemsa preparation reveals tumor cells or tumor-suspect cells, but the clinical examination, radiological studies, and standard laboratory tests provide no other evidence of a primary brain tumor or cerebral metastasis, i.e., in cases of pure meningeal carcinosis. If definite tumor cells are present according to the criteria described above, then a full complement of tumor marker studies should be undertaken to determine the tumor type and cell of origin (including a repeated lumbar puncture and testing of the newly obtained CSF for tumor markers), and an interdisciplinary tumor search is also indicated. When only tumor-suspect cells are found, tumor marker studies should begin with a limited search for the particular cell populations that are liable to be misdiagnosed as tumor cells (poorly differentiated, non-neoplastic precursor cells of meningeal origin, bone marrow cells, irritative forms of the lymphocytic and monocytic series, phagocytes, and non-neoplastic giant cell forms).

Immunocytochemical tumor marker studies are advisable when a primary CNS tumor is strongly suspected but not definitively diagnosed on clinical and radiological grounds, and tumor cells or tumor-suspect cells are found in the CSF sediment. A precise classification of these CSF cells is useful above all in determining the further treatment.

Immunocytochemical tumor marker studies of CSF cells are superfluous, in cases of biopsy-proven primary CNS tumors and brain metastases of known source (including generalized malignant lymphoma with CNS infiltrates), as well as in leukemia, which can generally be diagnosed and precisely classified with diagnostic testing of the blood, bone marrow, or both.

We conclude this general introductory section with an important remark: tumor cells or tumor-suspect cells may be present in the CSF in very small numbers, as only a few individual cells or small aggregates. The CSF cytologist will be able to diagnose a tumor (or at least a suspected tumor) from the CSF sediment even though the CSF cell count is not elevated above normal. Thus, whenever there is a question of a possible CNS tumor, a CSF sediment should always be prepared and meticulously examined, regardless of the cell count.

Fig. 5.1 Suspected tumor cells in a patient with grade I pilocytic astrocytoma, characterized by a round or lobulated nucleus with multiple nucleoli; a variable nuclear-to-cytoplasmic ratio; and a partially vacuolated, mostly acidophilic cytoplasm with basophilic areas. Compare the size of these cells with that of the two monocytes (left of center) and the lymphocyte (top). Immunocytochemical studies for tumor markers are indicated.

Fig. 5.2 Histologically confirmed grade II astrocytoma. The CSF cytological picture alone arouses no more than a remote suspicion of neoplasia. It is dominated by mononuclear reactive forms, probably originating from monocytes and/or ependymal structures. The nuclei are of variable shape with a single nucleolus; the cytoplasm is relatively large with some poorly demarcated areas, which are mostly acidophilic, but sometimes mildly basophilic at the periphery. For comparison, the figure also shows activated monocytes and a lymphocyte. A signet-ring cell is seen in the lower left of the figure. Follow-up examination of the CSF is indicated.

Fig. 5.3 Uniform cellular reaction in ventricular CSF obtained from an Ommaya reservoir, arousing the remote suspicion of astrocytoma in a 44-year-old man who had a histologically confirmed astrocytoma as a child. The figure shows mononuclear cells with relatively large nuclei, single nucleoli, and a strongly basophilic cytoplasm; these may represent cell types undergoing ependymal or endothelial transformation. Immunocytochemical tumor marker studies are indicated.

Fig. 5.4 A cluster of atypical cells in a patient with a histologically confirmed grade II astrocytoma. The nuclei and nucleoli are of variable size, and the basophilic cytoplasm has marked vesiculation. Immunocytochemical tumor marker studies are indicated.

Fig. 5.5 Markedly basophilic cell (tumor-suspect cell) with a large nucleus, multiple nucleoli, and vacuolated cytoplasm in a patient with grade II astrocytoma. Eosinophilic reactive pleocytosis with activated monocytes is seen. Immunocyto-chemical tumor marker studies are indicated.

Fig. 5.6 Cluster of tumor cells in grade III anaplastic astrocytoma, with variable configurations of cell nuclei, nucleoli, and cytoplasmic components. The predominantly basophilic but partly acidophilic cytoplasm has perinuclear vacuolation. There are accompanying pleocytosis and hemorrhage.

Fig. 5.7 Mononuclear, polyploid tumor cell in a patient with grade III astrocytoma. The nucleus has multiple nucleoli of variable size; the ring-shaped, strongly basophilic cytoplasm has numerous protuberances. Compare with the surrounding mononuclear cells, some of which are activated, and one of which is apparently beginning to phagocytose an erythrocyte. Accompanying hemorrhage and reactive pleocytosis with eosinophilic granulocytes are seen.

Fig. 5.8 Cluster of relatively uniformly shaped tumor cells in a patient with grade III astrocytoma. Some of the nuclei are polyploid; many nucleoli are seen, and the cytoplasmic component of the cells is markedly basophilic, although to a variable extent. A mitosis in anaphase/telophase is seen in the bottom right of the figure.

Fig. 5.9 Tumor cells in a patient with grade III astrocytoma, with conspicuously lobulated, sometimes coarsely structured nucleus. There is marked nuclear polychromasia. The cytoplasm is basophilic with perinuclear clearing. Adjacent to the tumor cells are possible “mini-tumor cells” or small cellular fragments (apoptotic nuclear fragments), some of which are ringed by cytoplasm.

Fig. 5.11 Cluster of tumor cells of varying size and marked uptake of stain in a patient with grade III astrocytoma. The arrow indicates a degenerating tumor cell of signet-ring type with vacuolated cytoplasm and intracytoplasmic nuclear fragments.

Fig. 5.12 Loose cluster of tumor cells in a patient with grade III astrocytoma, with a number of small cells and one highly polyploid giant tumor cell with a thin rim of markedly basophilic cytoplasm. Also present are degenerating cells and ghosts.

Fig. 5.13 Neoplastic meningitis in a patient with glioblastoma multiforme, with accompanying hemorrhage and relatively acute reactive pleocytosis. There are numerous polymorphic tumor cells with typical malignant features and variable affinity for stain, predominantly mononuclear; the nucleus has many nucleoli of variable size, and the cytoplasm of some of the cells displays peripheral vacuolation. Pathological mitosis is seen on the right. The granulocytes and erythrocytes provide an indication of the scale.

Fig. 5.14 A cluster of predominantly mononuclear tumor cells, some of which have a poorly defined border between the nucleus and the cytoplasm, in a patient with glioblastoma multiforme. Binucleated giant tumor cell with endocytogenesis, multiple nucleoli, and a vacuolated rim of cytoplasm. There is marked uptake of stain in the nucleus and cytoplasm. Pathological mitosis in prophase is seen (arrow).

Fig. 5.15 Mononuclear, relatively uniform tumor cells with variable affinity for stain and large nucleoli, in a patient with glioblastoma multiforme. Pathological mitosis in pro-/metaphase in a large tumor cell: chromosome deformation and reticulated cytoplasmic structures are easily seen. The granulocyte provides an indication of the scale. There is evidence of accompanying hemorrhage (degenerating erythrocytes).

Fig. 5.16 Relatively uniform, predominantly mononuclear tumor cells in a patient with glioblastoma multiforme. The nucleus is relatively well demarcated from the surrounding cytoplasm. In the center is a multinucleated, polyploid giant tumor cell that contains various structures, some of which are poorly defined. It has a thin rim of basophilic cytoplasm. A pathological mitosis is seen at the bottom right, with abnormal chromosomal division (early anaphase?). The erythrocyte provides an indication of the scale.

Fig. 5.17 Overview of relatively uniform, mononuclear tumor cells, some of which are in clusters, in a patient with glioblastoma multiforme. The nuclei are of varying size and shape, and the cytoplasm of variable extent and staining characteristics. Two mitoses in early metaphase are seen (arrowheads), as well as one mitosis in anaphase (arrow).

Fig. 5.18 Tumor cells of variable size and variable nuclear and cytoplasmic structure, some of which are in a small cluster, in a patient with glioblastoma multiforme. The cytoplasmic border is sharp in some cells, and irregular in others, with protuberances. The nuclei are round or oval, there are a few nucleoli, and intracytoplasmic chromatin fragments are seen (arrows). At the bottom is a polyploid giant cell with an irregular nucleus, cytoplasmic protuberances, and broken-off fragments of cytoplasm. A typical signet-ring cell is seen in the cluster (dotted arrow), as is a mitosis in anaphase (arrowhead). The single erythrocyte provides an indication of the scale.

Fig. 5.19 Tumor cells with three and four nuclei in a patient with a giant cell glioblastoma. The tumor cells have nuclei of variable size and nucleoli, and display evidence of amitoses (arrows). In some places, intracytoplasmic fragments of nuclear chromatin are seen. The cytoplasm has an irregular or smooth border, and is partly acidophilic and partly basophilic. The artifactual admixture of erythrocytes provides an indication of the scale.

Fig. 5.20 Highly polyploid tumor cell in a patient with a giant cell glioblastoma. The nucleus is in amitosis (constriction). The tumor cell below it is probably undergoing apoptosis. Artifactual admixture of blood and activated lymphocytes (arrows) is present.

Fig. 5.21 CSF cytological preparation in a patient with a giant cell glioblastoma showing a cluster of tumor cells of variable size, with a highly polymorphic nucleus and a narrow, strongly basophilic rim of cytoplasm. There are multinucleated giant tumor cells within and adjacent to the cluster (arrows). At the bottom, two signet-ring type cells with large vacuoles show degenerative changes in their nuclei and cytoplasm. The erythrocytes provide an indication of the scale.

Fig. 5.22 Mononuclear, highly polyploid giant tumor cell in a patient with a giant cell glioblastoma. The otherwise smooth nuclear membrane is notched at one point (early amitotic division?). The thin rim of cytoplasm is heavily stained and bears fringelike protuberances. A few erythrocytes are also seen.

Fig. 5.23 Highly polyploid giant tumor cell in tripolar amitosis, with nuclei of different sizes (partial separation from cell cluster), in a patient with a giant cell glioblastoma. The narrow rim of polychromatic cytoplasm is of variable depth, and its border is irregular in some areas. Adjacent to the giant tumor cell are other mononuclear and multinucleated tumor cells and a few erythrocytes.

Fig. 5.24 Loose cluster of tumor cells in a patient with a giant cell glioblastoma. Cells of variable size, some of which are separating from the cluster. Two mitoses in prophase are seen as well as a highly polyploid giant tumor cell in amitotic division (arrow) with thin, deeply stained, irregularly bordered cytoplasmic rim. The erythrocytes are shown for scale.

Fig. 5.25 Giant cell glioblastoma with polymorphic cells. The lumbar puncture was done while the patient was undergoing chemotherapy. A giant tumor cell is undergoing double amitosis. It has a large, partially fragmented cytoplasmic area with markedly basophilic periphery. Apoptotic cells are seen nearby (phagocytosis of nuclear fragments?).

Fig. 5.26 Giant tumor cell in apoptosis in a patient with giant cell glioblastoma. The erythrocyte provides an indication of the scale.

Fig. 5.27 Apoptosis of a giant tumor cell, and adjacent multinucleated giant cell, in a patient with glioblastoma. The erythrocyte and the monocyte indicate the scale.

Fig. 5.28 Probable primitive polar spongioblastoma of the cerebellum in a child. This is an independent, highly malignant disease entity (not to be confused with pilocytic astrocytoma). The illustration shows a loose isomorphic cluster of tumor cells, some of which are breaking away from the cluster. The nucleus is variably lobulated with many nucleoli and there is variable expression of malignant features. Some cells exhibit vesiculation of the cytoplasm. Note the nuclear fragmentation and apoptotic generation of extracellular nuclear fragments (arrow).

The current WHO classification includes the following types of ependymal tumor: myxopapillary ependymoma and subependymoma (grade I), ependymoma (grade II), and anaplastic ependymoma (grade III). Grade II ependymoma has four histopathological variants: cellular, papillary, clear cell, and tanycytic. Grade II ependymoma is distinguished from grade III anaplastic ependymoma by the absence of significant mitotic activity and other anaplastic features.

The cytoplasm of ependymoma cells occasionally contains glial filaments that can be revealed by a marker test for GFAP (which is not present in normal ependymal cells). This intermediate type between ependymoma and astrocytoma, called subependymoma, shows a low level of anaplasia. In general, the variants of grade II ependymoma are sometimes difficult to distinguish from low-grade astrocytic tumors, oligodendroglial tumors, and plexus papillomas, forcing the neuropathologist to resort to immunohistochemical marker profiles.

The neuropathological literature describes ependymoblastoma as another type of tumor that must be distinguished from anaplastic ependymoma. This type of tumor has a stem-cell character, is assigned WHO grade IV, and is considered to belong to the class of embryonal tumors.