Aberrant Localization of the Neuronal Class III [Beta]-Tubulin in Astrocytomas.

A Marker for Anaplastic Potential

Gliomas are the most common primary brain tumors in patients of all ages.[1-5] Most gliomas are of astrocytic origin (astrocytomas).[1-5] The most recent World Health Organization (WHO) classification (1993) separates astrocytic tumors into 2 main groups: the diffuse astrocytomas (WHO grades 2-4) and "other astrocytomas," the latter comprising a number of fairly distinct clinicopathologic entities, of which pilocytic astrocytoma is by far the most common.[4] This important distinction highlights the fact that pilocytic astrocytomas (WHO grade 1) and diffuse fibrillary astrocytomas (WHO grade 2) are fundamentally different entities.[1-6] Pilocytic astrocytomas are, for the most part, indolent neoplasms that only exceptionally undergo anaplastic (malignant) transformation or exhibit aggressive behavior.[1-7] The diffuse fibrillary astrocytomas, on the other hand, despite their well-differentiated, low-grade histologic appearance, frequently undergo malignant change with time and progress to high-grade lesions, anaplastic astrocytoma (WHO grade 3) or glioblastoma multiforme (WHO grade 4).[1-5] The prognosis of patients with high-grade astrocytomas remains poor, notwithstanding advances in surgical, radiation, and drug therapies, including the use of microtubule-acting compounds, such as paclitaxel (Taxol).[8-10]

The class III [Beta]-tubulin isotype ([Beta]III) is 1 of 6 [Beta]-tubulin isotypes expressed in mammals.[11] Its expression is contemporaneous with the earliest phase of neural differentiation and is neuron associated.[12-16] We have previously shown that this structural protein is expressed in neuronal/neuroblastic tumors, such as cerebellar medulloblastomas,[17-20] retinoblastomas,[21] central neurocytomas,[22] olfactory neuroblastomas,[23] sympathoadrenal neuroblastomas,[24] and adrenal pheochromocytomas,[25] as well as in neuronal/neuroblastic tumor cell lines.[26-29] Presently, [Beta]III is widely used as a neuronal marker in developmental biology and tumor pathology. However, [Beta]III is also expressed in nonneuronal tumors whose nontransformed, mature "cell-of-origin" analogues do not normally express this isotype. We have recently found that although the expression of [Beta]III in neuronal/neuroblastic tumors is differentiation dependent, its localization in certain nonneuronal neoplasms, such as small cell lung cancer, is associated with high-grade malignancy (anaplasia or "dedifferentiation").[30] To determine if the same pattern holds true for nonneuronal tumors of the central nervous system (CNS), we examined the immunoreactivity profile of [Beta]III in human astrocytomas. We now demonstrate that [Beta]III is a potential marker of high-grade astrocytomas, and its expression may also define incipient anaplastic phenotypes in diffuse fibrillary astrocytomas.

MATERIALS AND METHODS

Tissue Samples

Cases of adult (n = 55) and pediatric (n = 5) astrocytomas, obtained during a 10-year period (1989-1999), were retrieved from the files of the Department of Histopathology and Morbid Anatomy, The Royal London Hospital, Whitechapel, London, England, and the Department of Anatomic Pathology and Cytology, University of Patras Hospital, Rion, Patras, Greece. All specimens were originally diagnosed according to conventional histopathologic criteria based on the 1993 WHO Histological Typing of Tumors of the Central Nervous System.[4] The following common histologic types were selected for this study: pilocytic astrocytoma (WHO grade 1) (n = 8; age range, 2-19 years; median age, 12 years); diffuse fibrillary astrocytoma (WHO grade 2) (n = 18; age range, 1.1-64 years; median age, 38 years); anaplastic astrocytoma (WHO grade 3) (n = 4; age range, 30-47 years; median age, 44 years); and glioblastoma multiforme (WHO grade 4) (n = 30; age range, 3-73 years; median age, 58 years). Fifty-four tumors were supratentorial, and 6 were cerebellar (5 pilocytic astrocytomas and 1 diffuse fibrillary astrocytoma). Three of 8 pilocytic astrocytomas were supratentorial (1 diencephalic and 2 lobar/cerebral hemispheric). Optic gliomas, subependymal giant cell astrocytomas (SEGAs), pleomorphic xanthoastrocytomas (PXAs), mixed gliomas with oligodendroglial features (oligoastrocytomas), and gangliogliomas were excluded. All specimens were derived from wide tumor resections or debulking or from excisional biopsies: Stereotactic biopsy specimens were not included in this study. Of the 60 astrocytic tumors, 56 were primary excisions upon initial clinical presentation, and 4 had been previously treated with radiotherapy. None of the patients had received chemotherapy. Normal human CNS tissues from individuals of different ages (fetal to adult), obtained at autopsy, served as controls.

Tissue Fixation

All specimens were fixed in 10% neutral buffered formalin by immersion and processed conventionally for histology and immunohistochemistry. Sections 5 to 6 [micro]m in thickness were stained with hematoxylin-eosin for histologic evaluation and the remainder of the serial unstained sections were used for immunohistochemistry.

Antibodies

Two anti-class III [Beta]-tubulin antibodies (both produced by A.F. and commercially available through Covance, Richmond, Calif) were used: (1) the mouse monoclonal antibody TuJ1 (immunoglobulin [Ig] class G2a [IgG2a]) and (2) an affinity-purified rabbit antiserum specific for the same epitope as that recognized by monoclonal antibody TUJ1. Both antibodies were used at a dilution of 1:500. The staining pattern produced by the anti-[Beta]III antiserum is identical to that produced by TuJ1.[16,30] The production, purification, and characterization of these antibodies have been described previously.[13,30-32] All tumors examined in this study were also stained in parallel with commercially obtained mouse monoclonal antibodies to glial fibrillary acidic protein (GFAP) (clone 6F2; IgG1k; Dako Corporation, Santa Barbara, Calif; dilution, 1:100); synaptophysin (clone SVP-38; IgG1; Chemicon, Tamecula, Calif; dilution, 1:500); and Ki-67 nuclear antigen (clone NC-MM1; IgG2a; Novocastra, Newcastle-upon-Tyne, England, UK; dilution 1:200). In addition, an affinity-purified rabbit polyclonal antibody against the BM89 synaptic vesicle antigen (dilution, 1:20) was used as another neuronal marker for comparison.[33] The production, purification, and characterization of the BM89 polyclonal antibody have been described previously.[33] The BM89 antigen is an integral membrane glycoprotein with a molecular weight of 41 kd.[34] It exhibits substantial sequence homology with human synaptophysin (94.8% identity).[33] The BM89 antibodies cross-react with the L2/HNK-1 carbohydrate epitope expressed by members of a large family of glycoproteins.[34]

Immunohistochemistry

Immunoperoxidase studies were performed on deparaffinized sections of surgical and postmortem specimens according to the avidin-biotin complex (ABC) peroxidase method as previously described,[30] using rabbit IgG and mouse IgG ABC Elite Vectastain kits (Vector Laboratories, Burlingame, Calif) for the polyclonal and monoclonal antibodies, respectively. For TuJ1 or the anti-[Beta]III antiserum, no antigen retrieval was performed because no differences have been detected with respect to the distribution of immunoreactivity of either antibody after a microwave-based method of antigen retrieval.[30] Negative controls included normal rabbit IgG, nonspecific mouse ascites fluid (Becton-Dickinson, San Jose, Calif), or unrelated primary antibodies.

For double-labeling, the DAKO EnVision Doublestain System (K1395, Dako) was used with minor modification. Briefly, the sections first were deparaffinized in xylenes, rehydrated through a series of ethanols, and rinsed in Tris-buffered saline. Blocking for endogenous peroxidase was performed next, followed by sequential immunolabeling with primary mouse monoclonal antibody (TUJ1, 1:500) and secondary antibody (horseradish peroxidase conjugated), and the addition of the substrate chromogen diaminobenzidine (DAB). This sequence was followed by blocking with Doublestain blocking reagent; incubations with a second primary polyclonal antibody (rabbit anti-cow GFAP, N1506, prediluted by the manufacturer) and a secondary antibody (alkaline phosphatase conjugated); and the addition of the second substrate chromogen, 5-bromo-4-chloro-3-indolyl phosphate (BCIP)/ nitroblue tetrazolium (NBT). For controls, negative-control reagents supplied with the Doublestain kit were substituted for either the first primary antibody or the second primary antibody in the dual immunolabeling scheme. In addition, sections of human fetal telencephalon were used for positive controls for each antibody individually. The incubations for both primary and secondary antibodies were 30 minutes each. Substrate chromogens were incubated for ~10 minutes. All slides were counterstained with 0.1% nuclear fast red (Sigma Chemical Company, St Louis, Mo) for 1 minute, rinsed in Tris-buffered saline and double-distilled [H.sub.2]O, and coverslipped with Permount (Fisher Scientific, Pittsburgh, Pa).

Analysis of Staining and Statistical Methods

Histologic preparations were evaluated by a panel of 4 neuropathologists (C.D.K., L.D., J.F.G., and J.N.V.) Each member of the panel independently evaluated the specimens and assigned a histologic classification according to homogeneous criteria. In cases of disagreement, histologic typing was assigned by consensus at conference. Manual cell counting of labeled tumor cells was performed by 2 observers independently (C.D.K. and L.D.) Twenty nonoverlapping high-power fields (field magnification, x40) were evaluated, and the percentage of labeled tumor cells was calculated for each specimen and for each antibody. The total number of cells counted per specimen ranged from 501 to 1088 (median, 726). Interobserver agreement for both histologic classification and evaluation of immunohistochemical staining was within 15% ([Kappa] = 0.82).[35]

The cases were grouped into high-grade astrocytomas (consisting of anaplastic astrocytoma, WHO grade 3, and glioblastoma multiforme, WHO grade 4); diffuse fibrillary astrocytomas (including gemistocytic astrocytomas, WHO grade 2); and pilocytic astrocytomas (WHO grade 1). To assess the fraction of immunolabeled cells in specimens from each patient case, the labeling index (LI) was determined; this index was defined as the percentage of [Beta]III-positive (labeled) or Ki-67-positive cells out of the total number of tumor cells counted. In the case of Ki-67, labeled cells in tumor blood vessels/microvascular proliferation were excluded from the cell counts. The median LI (MLI) and interquartile range (IQR) of LIs were determined for the set of cases comprising each histologic grade using the UNIVARIATE procedure of the SAS package (SAS Institute, Cary, NC). The IQR is delimited by the 25th and 75th population percentiles. The statistical significance of differences in LIs between groups was examined with nonparametric statistical techniques using Kruskal-Wallis analysis of variance tests and the Wilcoxon rank sum post hoc tests. These analyses were carried out using the NON-PAR1WAY procedure of the SAS package (SAS Institute).

To investigate the existence of a relationship between [Beta]III and Ki-67 LIs and an ascending degree of malignancy, linear regression analysis was performed between the WHO tumor grades and the LIs for each of these markers.

RESULTS

[Beta]III Localization in Relation to Localization of GFAP and BM89 Synaptic Vesicle Antigen/Synaptophysin

A comparison of the MLIs for [Beta]III in each of the histologic grades is presented in Table 1 and Figure 1. The MLI for all 34 high-grade astrocytomas (anaplastic astrocytomas and glioblastomas) was 35% (IQR, 20%-47%). The MLI for glioblastomas was 34% (IQR, 17%-47%). [Beta]III was localized in 33 of 34 tumor specimens that were derived from both adult and pediatric patients. It was absent from a single specimen obtained during primary excision from a 71-year-old woman. In the remaining 33 cases, the range of [Beta]III LIs was 17% to 77%. When the small group of anaplastic astrocytomas (n = 4; MLI, 44%; IQR, 35%-56%) was split away from the glioblastomas (n = 30), no statistically significant differences were present between these 2 subgroups of high-grade astrocytomas.

[ILLUSTRATION OMITTED]

Table 1. [Beta]III Labeling Index(*)

                                      MLI, %          IQR, %
Tumor Group                   No.   ([dagger])   ([double dagger])

Pilocytic astrocytoma
  (WHO grade 1)                8    0(a,b,c,d)          0-0.5
Diffuse fibrillary
  astrocytoma (WHO grade 2)   18    4(a,e,f,g)         0.2-21
High-grade astrocytomas
(WHO grades 3 and 4)          34    35(b,e)             20-47
  Anaplastic astrocytoma
    (WHO grade 3)              4    44(c,f)             35-56
  Glioblastoma multiforme
    (WHO grade 3)             30    34(d,g)             17-47

                             Minimum            Maximum
Tumor Group                   LI, %              LI, %

Pilocytic astrocytoma
  (WHO grade 1)               0.0                 0.8
Diffuse fibrillary
  astrocytoma (WHO grade 2)   0.0                  33
High-grade astrocytomas
(WHO grades 3 and 4)          0.0([sections])      77
  Anaplastic astrocytoma
    (WHO grade 3)             28                   66
  Glioblastoma multiforme
    (WHO grade 3)             0.0([sections])      77

(*) [Beta]III indicates class III [Beta]-tubulin isotype; No., the
number of tumor specimens, each corresponding to a tumor case; MLI,
median labeling index; IQR, interquartile range; LI, labeling index;
and WHO, World Health Organization.

([dagger]) The letter (a) indicates pilocytic astrocytoma vs diffuse
fibrillary astrocytoma, P < .01; (b), pilocytic astrocytoma
vs high-grade astrocytomas, P < .0001; (c), pilocytic
astrocytoma vs anaplastic astrocytoma, P < .007; (d),
pilocytic astrocytoma vs glioblastoma multiforme, P <
.0001; (e), diffuse fibrillary astrocytoma vs high-grade
astrocytomas, P < .0001; (f), diffuse fibrillary astrocytoma vs
anaplastic astrocytoma, P < .004; (g), diffuse fibrillary astrocytoma
vs glioblastoma multiforme, P < .0001.

([double dagger]) The interquartile range is delimited by the 25th
and 75th population percentiles.

([sections] Observed only in a single case of glioblastoma
multiforme (see "Results").

In high-grade astrocytomas, [Beta]III-positive tumor cells exhibited a variety of morphologic appearances. Widespread, variably intense [Beta]III staining was present in the cytoplasm of overt astroglial phenotypes with multipolar fibrillary processes (Figure 2, a through f), in large "ganglioid" astroglial cells with prominent nucleoli and fibrillary cell processes (Figure 2, f), and in small anaplastic cells resembling "primitive" glioblasts (Figure 2, g). Typically, the localization of [Beta]III was diffuse within the individual tumor cells and was distributed equally in both the perikaryal cytoplasm and in major fibrillated cell processes (Figure 2, a through f). Both filamentous (Figure 2, b, d, and f) and granular patterns of localization (Figure 2, c and e [arrow]) were present. Occasional tumor cells in anaplastic astrocytomas and glioblastomas exhibited a partial or irregular localization with a predilection for the periphery of the cell body (Figure 3, a) (vide infra). Neoplastic [Beta]III-positive cells formed either large sheets or small clusters and tended to aggregate around tumor blood vessels (Figure 2, c, h, and i; Figure 4, a, c, and e) or in the vicinity bordering tumor necrosis (Figure 2, g). In contrast, tumor-associated blood vessels, exhibiting florid angioproliferative changes and/or stromal proliferations of smooth muscle cells, were consistently [Beta]III negative (Figure 2, c, h, and i; Figure 4, a, c, and e).

[ILLUSTRATIONS OMITTED]

Co-localization of GFAP and [Beta]III within individual tumor cells was demonstrated by double-labeling (Figure 3, a and b). The 2 chromogenic reaction products were either differentially distributed (Figure 3, a) or intermingled within the cytoplasm of neoplastic astrocytes (Figure 3, b). The former pattern was exemplified in subpopulations of tumor cells of anaplastic astrocytomas and glioblastomas in which [Beta]III immunoreactivity exhibited a rimlike peripheral distribution, whereas GFAP was localized in the residual cytoplasm (Figure 3, a). Most [Beta]III-positive tumor cells were also GFAP positive in immediately adjacent sections (Figure 4, a through f).

Notwithstanding the small number of postirradiation cases included in this study, there were no significant differences in the number of [Beta]III-positive cells between tumor specimens from primary excisions and those that were extirpated after radiation therapy.

In sharp contrast to the [Beta]III immunoreactivity noted in high-grade astrocytomas, there was a paucity or lack of [Beta]III immunoreactivity in the low-grade pilocytic astrocytomas (Table 1, Figure 5, a). The MLI for pilocytic astrocytomas was 0% (IQR, 0%-0.5%) (P [is less than] .0001 vs high-grade astrocytomas). In diffuse fibrillary astrocytomas, the distribution of [Beta]III immunoreactivity was also substantially less than that in high-grade astrocytomas (MLI, 4%; IQR, 0.2%-21%) (P [is less than] .0001) (Figure 1). However, in diffuse fibrillary astrocytomas, there was considerable [Beta]III staining heterogeneity among different tumors (range, 0%-33%), the highest LI corresponding to the gemistocytic variant. Variability of [Beta]III-positive cells was also common within individual tumors (intratumoral staining heterogeneity). In 5 of 18 cases, the LIs (values [is greater than] 20%) overlapped the LIs of high-grade astrocytomas (Figure 1). Some clusters of fibrillated and gemistocyte-like cells in diffuse astrocytomas were [Beta]III positive, whereas others were [Beta]III negative (Figure 5, b through d). The [Beta]III-positive cells were morphologically indistinguishable from the predominantly [Beta]III-negative cells (Figure 5, b through d).

[ILLUSTRATION OMITTED]

In the normal CNS tissue from patients of different ages, [Beta]III localization was solely neuron specific, as described previously[15] (data not shown).

No immunoreactivity was detected in astrocytomas using either the polyclonal antibody to BM89 synaptic antigen/synaptophysin or the monoclonal antibody SVP-38 against synaptophysin (Table 2). The distribution of staining of BM89 and SVP-38 in normal CNS was neuron specific (data not shown), exhibiting predominantly neuropil and perisomatic patterns, as previously described.[34] Perikaryal localization of BM89 was also detected in subpopulations of rhombencephalic neurons. The BM89 synaptic antigen/synaptophysin staining was invariably present in the entrapped or infiltrated neuropil by the neoplastic process (data not shown).

Table 2. Comparison of [Beta]III, GFAP, and Synaptophysin
Immunoreactivity Profiles in Astrocytic vs Neuronal/
Neuroblastic Tumors of the CNS(*)

Tumor Group                     [Beta]III

Astrocytomas
  Pilocytic astrocytoma
    (WHO grade 1)               -
  Diffuse fibrillary
    astrocytoma (WHO grade 2)   +([dagger])
  High-grade astrocytoma
    (anaplastic astrocytomas
    and glioblastoma
    multiforme; WHO grades
    3 and 4)                    ++([dagger])

Neuronal/neuroblastic tumors
  Medulloblastoma.[17-20]       ++/+++([double dagger])
  Retinoblastoma[21]            ++/+++([sections])
  Central neurocytoma[22]       ++/+++([parallel])

Tumor Group                     GFAP                   Synaptophysin

Astrocytomas
  Pilocytic astrocytoma
    (WHO grade 1)               ++                     -
  Diffuse fibrillary
    astrocytoma (WHO grade 2)   +++                    -
  High-grade astrocytoma
    (anaplastic astrocytomas
    and glioblastoma
    multiforme; WHO grades
    3 and 4)                    ++/+++([paragraph])    -

Neuronal/neuroblastic tumors
  Medulloblastoma.[7-20]        - (+ reactive
                                  astrocytes;
                                  exceptionally,
                                  + neoplastic glia)   ++/+++
  Retinoblastoma[21]            - (+ reactive
                                  astrocytes)          ++/+++
  Central neurocytoma[22]       - (+ reactive
                                  astrocytes)          ++/+++

(*) [Beta]III indicates class III [Beta]-tubulin isotype; GFAP,
glial fibrillary acidic protein, CNS, central nervous system; and
WHO, World Health Organization. Rating of the distribution of
immunoreactivity: -, absent; +, focal, scanty/sparse; ++, moderate,
heterogenerous; and +++, widespread and dense. (See Table 1 for
[Beta]III median labeling indices, interquartile ranges, and full
ranges of labeling indices in astrocytic tumors.)

([dagger]) Intratumoral staining heterogeneity independent of
morphologic phenotype.

([double dagger]) Neuronal differentiation dependent; associated
with neoplastic neuritogenesis (Homer Wright rosettes, "pale islands"
of desmoplastic medulloblastomas).

([sections]) Neuronal differentiation dependent (including
Flexner-Wintersteiner rosettes and "fleurettes.")

([parallel]) Neuronal differentiation dependent.

[paragraph]) Co-localization of 13111 and GFAP in neoplastic
astrocytes.

[Beta]III Immunoreactivity in Relation to Ki-67 Nuclear Antigen Immunoreactivity

A comparison between the LIs of [Beta]III and Ki-67 in a sample of 25 astrocytomas representative of all tumor grades is presented in a scatter plot graph (Figure 6). The Ki-67 MLI for high-grade astrocytomas, WHO grades 3 and 4 combined, was 24% (IQR, 17%-25%) (P [is less than] .002 vs diffuse fibrillary astrocytomas; P [is less than] .0001 vs pilocytic astrocytomas). The Ki-67 MLIs were 8% and 24% (IQR, 18%-25%) for anaplastic astrocytomas and glioblastomas, respectively (P [is less than] .001 vs diffuse fibrillary astrocytomas). In contrast, the Ki-67 MLI was 5% (IQR, 3%-8%) for diffuse fibrillary astrocytoma and 2% (IQR, 0.3%-5%) for pilocytic astrocytoma.

[ILLUSTRATION OMITTED]

Linear regression analysis revealed the existence of a highly significant relationship between [Beta]III and Ki-67 LIs and an ascending degree of malignancy. However, this relationship was stronger for Ki-67 than for [Beta]III (Ki-67, P [is less than] .0001; [Beta]III, P [is less than] .006).

COMMENT

In this report, we describe the cellular distribution of the neuron-associated [Beta]III in 60 surgically excised common astrocytic gliomas, and we provide new evidence that aberrant expression of this neuronal cytoskeletal protein is a molecular signature of malignancy in astroglial tumors. We have examined only a series of common astrocytic tumors with clear-cut histologic type, grade, and supporting clinicopathologic features. To avoid sampling error and to adequately assess [Beta]III immunoreactivity in terms of intratumoral staining heterogeneity, the study was restricted to relatively large resection and extirpation specimens representative of all salient histologic features typifying each tumor entity.

We demonstrated that [Beta]III immunoreactivity, which is not normally present in developing or mature astrocytes,[13-15,21,36-41] is significantly increased in high-grade astrocytic tumors, notably, anaplastic astrocytomas and glioblastomas (WHO grades 3 and 4, respectively). In contrast, [Beta]III immunoreactivity was present to a lesser extent in diffuse fibrillary astrocytomas (WHO grade 2) and was rarely detectable in pilocytic astrocytomas (WHO grade 1). A statistically significant difference in [Beta]III LIs was demonstrated between grade 2 and grades 3 and 4 astrocytic tumors. On the other hand, no statistically significant difference was demonstrated between the grade 3 and grade 4 subgroups of high-grade astrocytomas (notwithstanding the very small number of grade 3 astrocytomas included in this series). Thus, in the context of astrocytic gliomas, [Beta]III staining mirrors the degree of aggressiveness of the 4 types of common tumors. The most aggressive tumors, glioblastomas and anaplastic astrocytomas, consistently exhibited the most widespread staining, whereas the least aggressive of the 4 astrocytoma entities, the pilocytic astrocytomas, exhibited little or no staining. No statistically significant differences in [Beta]III LIs were present in tumor specimens derived from primary excisions versus those that were extirpated following radiation therapy. However, the number of postirradiation cases included in this study was very small (n = 4) and not significant for statistical analysis, thus warranting further studies in this regard.

To further corroborate that [Beta]III is expressed in astrocytomas according to an ascending gradient of malignancy, we investigated the existence of a relationship between [Beta]III and Ki-67 LIs and malignancy. In addition to clinical parameters (ie, patient's age, Karnofsky score) and traditional histopathologic parameters, Ki-67 is currently the most widely used cell proliferation marker for the prognostic and predictive assessment of astrocytic gliomas because it correlates well with histologic malignancy and distinguishes grade 2 from grade 3 tumors.[42-45] Consistent with previous reports,[42-45] we found a significant difference between the Ki-67 LIs of diffuse fibrillary astrocytomas (WHO grade 2) and high-grade astrocytomas (WHO grades 3 and 4) in the present study. However, the NC-MM1 monoclonal antibody yielded consistently higher MLIs for each astrocytoma tumor group, as compared with the mean LIs reported in previous studies using the monoclonal antibody MIB-1.[43-45] No significant overlap was noted between Ki-67 LIs of high-grade and lesser-grade (WHO grade 2) astrocytomas. A variability of Ki-67 immunostaining in glioblastomas has previously been reported[46]; this variability may result from the use of different Ki-67 equivalent antibodies, the use of manual staining versus automated immunohistostainers, and the use of frozen sections versus paraffin-embedded sections.

Our results indicate that a highly significant, grade-dependent relationship exists for both [Beta]III and Ki-67 immunoreactivity and degree of malignancy, although the relationship is stronger for Ki-67 ([Beta]III, P [is less than] .006; Ki-67, P [is less than] .0001). Thus, [Beta]III is an additional marker of potential diagnostic significance in the context of high-grade astrocytomas. That said, in some high-grade astrocytoma resection specimens, there are areas in which tumor cells are [Beta]III negative. Consequently, it is still possible that a stereotactic biopsy could lead to a false-negative result. This may be exemplified in the single instance of glioblastoma in this series (a primary excision in a 71-year-old woman) in which [Beta]III immunoreactivity was not detectable, even though the tumor exhibited widespread GFAP staining in an immediately adjacent section.

The pathogenesis of human gliomas embodies the accumulation of multiple genetic alterations, which include chromosomal derangements and gene mutations, causing extensive changes in the expression of proteins involved in the regulation of cell proliferation and signal transduction.[47] To this end, the relationship between tumor aggressiveness, anaplastic phenotype, and [Beta]III immunoreactivity becomes most important with respect to the heterogeneous group of diffuse fibrillary astrocytomas (WHO grade 2). These lesions may be viewed as a biologic continuum arising as "low-grade," well-differentiated, slow-growing tumors, but culminating in many instances in high-grade malignancy through successive (albeit hitherto undefinable) steps.[5] Interestingly, certain diffuse fibrillary astrocytomas undergo a more rapid malignant transformation, whereas others are considerably more indolent.[5,48] To date, there are no molecular markers to identify potentially, or imminently, anaplastic phenotypes within the diffuse fibrillary astrocytoma group. It is possible that the relatively small number of [Beta]III-positive cells in diffuse fibrillary astrocytomas may represent subpopulations of more aggressive cells heralding anaplastic change in these tumors, although this question cannot be answered in the present study.

The class III [Beta]-tubulin isotype differs from the other tubulin isotypes with respect to its amino acid sequence[49] and posttranslational modifications, which include phosphorylation.[32,50-52] During development, [Beta]III expression appears to be neuron specific and the earliest lineage-specific marker protein for neurons.[22-16] The immunocytochemical localization of [Beta]III in immature neuroepithelial cells, including neural precursors, is widely construed as evidence of neuronal lineage differentiation.[12-16,36-41,52] This differentiation-dependent, neuronal lineage-associated expression of [Beta]III is recapitulated in neuronal/neuroblastic tumors.[17-25] On the other hand, [Beta]III has been detected previously by immunocytochemistry and immunoblotting in the U-251 MG human glioblastoma cell line[53] but is absent in the rat C6 glioma line.[54] In a similar vein, we have recently demonstrated that unlike the presence of [Beta]III in neuronal/neuroblastic tumors, the presence of [Beta]III in certain nonneuronal neoplasms, such as in epithelial neuroendocrine tumors of the lung, is associated with subpopulations of tumor cells in aggressive and extensively metastasizing tumors (ie, small cell lung cancer).[30] Deregulation of [Beta]III expression in neoplasia may reflect loss of epigenetic control, and thus "dedifferentiation" toward an immature precursor-like phenotype or phenotypes marking potentially anaplastic tumor subclones.

A multifold increase of [Beta]III transcript expression has been shown in nonneuronal tumors, such as non-small cell lung cancer, and in ovarian and prostate cancer cell lines, particularly in those exhibiting resistance to a class of antimicrotubule (microtubule-polymerizing) drugs, such as estramustine and paclitaxel (Taxol).[55-59] Paclitaxel and estramustine bind to [Beta]-tubulin causing microtubule polymerization, which blocks mitosis by kinetic stabilization of spindle microtubules.[60] Microtubules composed of [Beta]III are considerably less sensitive to the effects of Taxol as compared with microtubules assembled from unfractionated tubutin.[61] Potentially, this decreased sensitivity could confer a survival advantage to tumor cells exposed to microtubule-targeting drugs, such as Taxol and related microtubule-polymerizing compounds.[62] In tumor cell lines, 2 fundamental mechanisms of resistance to antimicrotubule agents have been described: (1) overexpression of the multiple drug resistance (MDR1) gene and (2) alterations in the cellular microtubule proteins, including mutations and/or differences in expression of the tubulin genes.[63]

Taxol alkaloid compounds such as paclitaxel are currently being explored as salvage chemotherapeutic agents in gliomas, for they appear to potentiate the effects of radiation therapy.[8-10] Because of their radiosensitizing effect and the overall lack of side effects (aside from immunologic toxicity), these compounds hold promise in the adjuvant chemotherapy of gliomas. However, the modest or suboptimal therapeutic response produced by these agents in malignant and recurrent gliomas points to the existence of Taxol-resistant phenotypes, which have been hitherto molecularly undefinable. As the binding sites of Taxol and related compounds have been delineated on the [Beta]-tubulin subunit, [Beta]III has become a potential marker for Taxol chemoresistance in gliomas, much as it is in tumors of the lung, prostate, and ovary.[55-59] In this regard, given the spread of [Beta]III LIs in glioblastomas, it remains to be determined whether those tumors with lower LI values may be more amenable to chemotherapy with Taxol compounds. Also, this raises the possibility for a potential therapeutic use of antisense oligonucleotides to [Beta]III as modulators capable of sensitizing resistant tumor cells to taxanes.[59] Thus, [Beta]III has potential implications in emerging therapies in which tumor-abrogating specificity is targeted at microtubule binding.

[Beta]III in Relation to Other Neuronal Markers in Astrocytomas

Notwithstanding a plethora of immunophenotypic studies on gangliogliomas and related putative glioneuronal neoplasms, the cellular distribution of neuronal markers in common astrocytomas remains underinvestigated. To a large extent, this reflects the widely held belief that the presence of a neuronal protein in a neoplastic cell denotes neuronal differentiation. Among the most studied neuronal cytoskeletal proteins in astrocytomas is neurofilament protein (NF).[64-66] Neurofilament-like immunoreactivities in astroglial tumors are unusually frequent, although published data differ considerably among studies. This variability may be due to the use of a wide range of anti-NF antibodies and the lack of systematic cell counts and statistical analyses. In this regard, certain well-characterized anti-NF monoclonal antibodies exhibit increased immunoreactivities in anaplastic gliomas.[65] This trend is analogous to our findings on [Beta]III immunoreactivity. Tlhyama et al[66] have previously demonstrated coexpression of low molecular weight NF and GFAP in human glioma cell lines. These authors have postulated that the presence of NF in neoplastic glial cells may either reflect aberrant expression or indicate bipotentiality.[66]

In contrast to the phenotypic spectrum of NE the spectrum of synaptophysin in CNS tumors is more restricted to bona fide neuronal/neuroblastic tumors,[2,3,5,17,19,21,22,67] the neuronal component of mixed glioneuronal neoplasms,[67] and/or subpopulations of neoplastic cells in certain cases of oligodendrogliomas.[68-70] The present study demonstrates that in contrast to [Beta]III, BM89 synaptic vesicle antigen/synaptophysin is absent in a wide range of ordinary astrocytic gliomas. The latter finding is in agreement with the previously reported findings by Miller and coworkers[67] with respect to synaptophysin. This observation has potential implications for the diagnosis of putative mixed glioneuronal tumors, insofar as the detection of [Beta]III in tumor cells is not tantamount to neuronal differentiation in the absence of synaptophysin staining (Table 2).

Ganglionic differentiation and divergent glioneuronal differentiation have been construed on the basis of localization of neuronal proteins in PXAs[71,72] and SEGAs.[73] Given the ambiguous glial and neuronal character of these 2 tumor types, the localization of [Beta]III raises important theoretical and practical questions. As exemplified in a case of PXA with a concomitant gangliogliomatous component,[72] PXAs may be part of a nosologic spectrum of "developmental neoplasms," overlapping with gangliogliomas, "dysembryoblastic neuroepithelial tumors" and desmoplastic gangliogliomas.[71,72] Consequently, these tumors may be fundamentally mixed glioneuronal neoplasms in which the presence of [Beta]III, especially in conjunction with synaptophysin, may reflect true divergent neuronal differentiation.

The localization of [Beta]III in subpopulations of tumor cells of certain human SEGAs of tuberous sclerosis has been interpreted as evidence of "divergent neuronal differentiation."[73] By the same token, [Beta]III is absent in the benign subependymal astrocytic hamartomas in adult Eker rat carriers of the genetic locus TSC2 (16p13) of tuberous sclerosis (TSC2 [+ or -]).[74] In light of the findings of the present study, the localization of [Beta]III in SEGAs warrants further explanation insofar as these are inherently low-grade tumors (WHO grade 1).

Based on the previously published report by Lopes et al[73] the nature of [Beta]III-positive phenotypes in SEGAs remains unclear. These phenotypes remain poorly defined because there are no data concerning [Beta]III LIs or the colocalization of [Beta]III/synaptophysin and GFAP/S100 protein/vimentin in individual tumor cells. The [Beta]III-positive phenotypes in SEGAs may thus represent (1) divergent neuronal phenotypes corresponding to transformed bipotential precursor cells of the subependymal matrix displaying either neuronal ([Beta]III-positive) or glial (GFAP-positive) differentiation,[73] (2) entrapped remnants of dysgenetic, incompletely differentiated neuronal cells resulting from deranged ontogeny and disrupted migration of neuronal precursors of the subependymal germinal matrix, or (3) transformed glial cells in which there is anomalous induction of [Beta]III that is not associated with cell proliferation (ie, in the unique setting of this dysgenetic tumor). In our view, the nature of [Beta]III-positive cells in SEGAs requires further elucidation.

Differential Distribution of [Beta]III in Neuronal/Neuroblastic Versus Astroglial Tumors

As compared with [Beta]III staining in astrocytic gliomas, [Beta]III staining is far greater and substantially more homogeneous in neuronal/neuroblastic tumors, where it is differentiation dependent and present consistently and contemporaneously with ensuing neoplastic neuritogenesis.[17-26] This is best exemplified in the reticulin-free "pale islands" of desmoplastic medulloblastomas, in which robust and widespread [Beta]III immunoreactivity, in conjunction with immunostaining for other neuronal markers such as synaptophysin, is an expression of neuronal differentiation and is accompanied by a low growth potential.[17-20,75-77] In contrast, the poorly differentiated, actively proliferating neoplastic neuroblasts in the surrounding reticulin-rich/ solid areas of tumor exhibit lesser degrees of [Beta]III labeling[18,20] Moreover, neuronal tumors are GFAP negative, aside from reactive or stromal astroglial proliferations.[17,20,21] Interestingly, nonneoplastic (reactive) astroglial cells in medulloblastomas are GFAP positive and [Beta]III negative.[17,20] On the other hand, astrocytic tumors are, on the whole, GFAP positive and synaptophysin negative; as shown in the present study, these tumors exhibit increasing numbers of [Beta]III-positive tumor cells as a reflection of histologic malignancy as opposed to a differentiating state. Because astrocytic tumors of all histologic grades are virtually homogeneous with respect to GFAP immunoreactivity, subpopulations of [Beta]III-positive cells are also GFAP-positive. A comparison of [Beta]III, GFAP, and synaptophysin immunoreactivities in astrocytic versus CNS neuronal/neuroblastic tumors is presented in Table 2.

The present study indicates a differential expression of [Beta]III in neuronal/neuroblastic tumors and astroglial tumors. This differential expression suggests a possible differential use of this structural protein in the context of growth and differentiation among different neuroepithelial tumors. Thus, in neuronal/neuroblastic tumors, [Beta]III localization mirrors morphogenetic events in the direction of maturation (reflecting a differentiation-dependent gradient), whereas in astrocytomas, increased [Beta]III LIs are associated with a tendency toward high-grade malignancy (anaplasia). Consistent with this observation, [Beta]III immunoreactivity is present in the neuronal but not in the innocuous (low-grade) astroglial component of gangliogliomas.[78]

In summary, this study demonstrates that under certain neoplastic conditions, [Beta]III is not restricted only to neuronal phenotypes, thus warranting cautious interpretation of [Beta]III-positive cells in the differential diagnosis of brain tumors. Our findings underscore the fact that tumor phenotypes do not necessarily follow developmental rules, and that their interpretation should be based on a critical assessment of tumor morphology in conjunction with other well-characterized cell lineage-associated markers. As such, [Beta]III staining should not be construed by itself as a priori evidence of "divergent" neuronal differentiation in neuroepithelial tumors that are otherwise phenotypically astroglial.

We thank Professor Sverre J. Murk, MD, PhD, Department of Pathology, The Gade Institute, University of Bergen, Haukeland Hospital, Bergen, Norway, and Elias Perentes, MD, DMSc, Preclinical Safety, Novartis Pharma, Basel, Switzerland, for their constructive comments. The technical contributions by Mr Ed Browning (Royal London Hospital) and Mrs Marra Vaynshteyn (St Christopher's Hospital for Children) are gratefully acknowledged. This research was supported by 2 grants from the National Institutes of Health: a Research Supplement for Individuals with Disabilities, National Institute for Neurological Disorders and Stroke, 3 PO1 NS36466-03 S1 (Dr Katsetos) and 3 PO1 NS36466 (Dr Khalili).

References

[1.] Russell DS, Rubinstein U. Pathology of Tumours of the Nervous System. 5th ed. London, England: Arnold; 1989:1-57, 83-350.

[2.] Lantos PL, Kleihues P, VandenBerg SV, Lopes MBS. Tumours of the central nervous system. In: Lantos PL, Graham DI, eds. Greenfield's Neuropathology, 6th ed. London, England: Arnold; 1997:583-879.

[3.] Schiffer D. Brain Tumors. Biology, Pathology and Clinical References. Berlin, Germany: Springer-Verlag; 1997.

[4.] Kleihues P, Burger PC, Scheithauer BW. Histological Typing of Tumors of the Central Nervous System. Berlin, Germany: Springer-Verlag; 1993.

[5.] Burger PC, Scheithauer BW. Tumors of the Central Nervous System. Washington, DC: Armed Forces Institute of Pathology; 1994:2-7, 25-161. Atlas of Tumor Pathology; 3rd series, fascicle 10.

[6.] Katsetos CD, Krishna L. Lobar pilocytic astrocytomas of the cerebral hemispheres: I. Diagnosis and nosology. Clin Neuropathol. 1994;13:295-305.

[7.] Tihan T, Fisher PG, Kepner IL, et al. Pediatric astrocytomas with monomorphous pilomyxoid features and less favorable outcome. J Neuropathol Exp Neurol. 1999;58:1061-1068.

[8.] Prados MD, Schold SC, Spence AM, et al. Phase II study of paclitaxel in patients with recurrent glioma. J Clin Oncol. 1996;14:2316-2321.

[9.] Chang SM, Kuhn JG, Rizzp J, et al. Phase I study of paclitaxel in patients with recurrent malignant glioma: a North American Brain Tumor Consortium report. J Clio Oncol. 1998;16:2188-2194.

[10.] Wehbe T, Glantz M, Choy H, et al. Histologic evidence of a radiosensitizing effect of taxol in patients with astrocytomas. J Neurooncol. 1998;39:245-251.

[11.] Sullivan KF, Cleveland DW. Identification of conserved isotype-defining variable region sequences for four vertebrate [Beta]-tubulin polypeptides classes. Proc Natl Acad Sci U S A. 1986;83:4327-4331.

[12.] Moody SA, Quigg MS, Frankfurter A. The development of the peripheral trigeminal system in the chick revealed by an isotype-specific anti-[Beta]-tubulin monoclonal antibody. J Comp Neurol. 1989;279:567-580.

[13.] Lee MK, Tuttle JB, Rebhun LI, Frankfurter A. The expression and posttranslational modification of a neuron-specific [Beta]-tubulin isotype during chick embryogenesis. Cell Motil Cytoskeleton. 1990;17:118-132.

[14.] Easter SS, Ross ES, Frankfurter A. initial tract formation in the mouse brain. J Neurosci. 1993;13:285-299.

[15.] Katsetos CD, Frankfurter A, Christakos S, Mancall EL, Vlachos I, Urich H. Differential localization of class III [Beta]-tubulin isotype ([Beta]III) and calbindin-[D.sub.28k] defines distinct neuronal types in the developing human cerebellar cortex. J Neuropathol Exp Neurol. 1993;52:655-666.

[16.] Katsetos CD, Karkavelas G, Herman MM, et al. Class III [Beta]-tubulin isotype in the adrenal medulla. I. Localization in the developing human adrenal medulla. Anat Rec. 1998;250:343-349.

[17.] Katsetos CD, Herman MM, Frankfurter A, et al. Cerebellar desmoplastic medulloblastomas. A further immunohistochemical characterization of the reticulin-free pale islands. Arch Pathol Lab Med. 1989;113:1019-1029.

[18.] Katsetos CD, Krishna L, Frankfurter A, et al. Cytomorphological scheme of differentiating neuronal phenotypes in cerebellar medulloblastomas based on immunolocalization of class III [Beta]-tubulin isotype ([Beta]III) and the proliferating cell nuclear antigen (PCNA)/cyclin. Clin Neuropathol. 1995;14:172-181.

[19.] Maraziotis T, Perentes E, Karamitopoulou E, et al. Neuron-associated class III [Beta]-tubulin isotype, retinal S-antigen, synaptophysin and glial fibrillary acidic protein in human medulloblastomas: a clinicopathological analysis of 36 cases. Acta Neuropathol (Bed). 1992;84:355-363.

[20.] Katsetos CD, Burger PC. Medulloblastoma. Semin Diago Pathol. 1994;11: 85-97.

[21.] Katsetos CD, Frankfurter A, Herman MM, Uffer S, Perentes E, Rubinstein LJ. Neuron-associated class III [Beta]-tubulin isotype, microtubule-associated protein 2 and synaptophysin in human retinoblastomas in situ: further immunohistochemical observations on Flexner-Wintersteiner rosettes. Lab Invest. 1991;64:45-54.

[22.] Hessler RB, Lopes MB, Frankfurter A, Reidy J, VandenBerg SR. Cytoskeletal immunohistochemistry of central neurocytomas. Am J Surg Pathol. 1992;16: 1031-1038.

[23.] Frierson HF, Ross GW, Mills SE, Frankfurter A. Olfactory neuroblastomas: additional immunohistochemical characterization. Am J Clin Pathol. 1990;94: 547-553.

[24.] Katsetos CD, Karkavelas G, Frankfurter A, et al. The stromal Schwann cell during maturation of peripheral neuroblastomas: immunohistochemical observations with antibodies to the neuronal class III [Beta]-tubulin isotype ([Beta]III) and S-100 protein. Clin Neuropathol. 1994;13:171-180.

[25.] Karkavelas G, Katsetos CD, Geddes JF, et al. Class III [Beta]-tubulin isotype in the adrenal medulla. II Localization in human primary pheochromocytomas. Anat Rec. 1998;250:344-350.

[26.] Herman MM, Perentes E, Katsetos CD, et al. Neuroblastic differentiation potential of the human retinoblastoma cell lines Y-79 and WERI-Rb1 maintained in an organ culture system. An immunohistochemical, electron microscopic and biochemical study. Am J Pathol. 1989;134:115-132.

[27.] Gass P, Frankfurter A, Katsetos CD, Herman MM, Donoso LA, Rubinstein LJ. Expression of neuron-associated microtubule protein h[Beta] 4 and MAP2 by the human retinoblastoma cell line WERI-Rb1: a comparative immunoblot and immunohistochemical study. Ophthalmic Res. 1990;22:57-66.

[28.] Vinores SA, Herman MM, Katsetos CD, May EE, Frankfurter A. Neuron-associated class III [Beta]-tubulin, tau and MAP2 in the D-283 Med cell line and in primary explants of human medulloblastoma. Histochem J. 1994;26:678-685.

[29.] Katsetos CD, Herman MM, Balin BJ, et al. Class III [Beta]-tubulin isotype ([Beta]III) in the adrenal medulla. III. Differential expression of neuronal and glial antigens identifies two distinct populations of neuronal and glial-like (sustentacular) cells in the PC12 rat pheochromocytoma cell line maintained in a Gelfoam matrix system. Anat Rec. 1998;250:351-366.

[30.] Katsetos CD, Kontogeorgos G, Geddes JF, et al. Differential distribution of the neuron-associated class III [Beta]-tubulin in neuroendocrine lung tumors. Arch Pathol Lab Med. 2000;124:535-544.

[31.] Lee MK, Rebhun LI, Frankfurter A. Posttranslational modification of class III [Beta]-tubulin. Proc Natl Acad Sci U S A. 1990;87:7195-7199.

[32.] Alexander JE, Hunt DF, Lee MK, et al. Characterization of posttranslational modifications in neuron specific class III [Beta]-tubulin by mass spectrometry. Proc Natl Acad Sci U S A. 1991;88:4685-4689.

[33.] Gaitanou M, Mamalaki A, Merkouri E, Matsas R. Purification and cDNA cloning of mouse BM89 antigen shows that it is identical with the synaptic vesicle protein synaptophysin. J Neurosci Res. 1997;48:507-514.

[34.] Mercouri E, Matsas R. Monoclonal antibody BM89 recognizes a novel cell surface glycoprotein of the L2/HNK-1 family in the developing mammalian nervous system. Neuroscience. 1992;50:53-68.

[35.] Altman DG. Practical Statistics for Medical Research. London, England: Chapman and Hall; 1991.

[36.] Menezes JRL, Luskin MB. Expression of neuron-specific tubulin defines a novel population in the proliferative layers of the developing telencephalon. J Neurosci. 1994;14:5399-5416.

[37.] Gates MA, Thomas LB, Howard EM, et al. Cell and molecular analysis of the developing and adult mouse subventricular zone of the cerebral hemispheres. J Comp Neurol. 1995;361:249-266.

[38.] Luskin MB, Zigova T, Soteres BJ, Stewart RR. Neuronal progenitor cells derived from the anterior subventricular zone of the neonatal rat forebrain continue to proliferate in vitro and express a neuronal phenotype. Mol Cell Neurosci. 1997;8:351-366.

[39.] Peretto P, Merighi A, Fasolo A, Bonfanti L. Glial tubes in the rostral migratory stream of the adult rat. Brain Res Bull. 1997;42:9-21.

[40.] Doetsch F, Garcia-Verdugo JM, Alvarez-Baylla A. Cellular composition and three-dimensional organization of the subventricular germinal zone in the adult mammalian brain. J Neurosci. 1997;17:5046-5061.

[41.] Weickert CS, Webster MJ, Colvin SM, et al. Localization of epidermal growth factor receptors and putative neuroblasts in human subependymal zone. J Comp Neurol. 2000;423:359-372.

[42.] Ellison DW, Steart PV, Baterman AC, Pickering RM, Palmer JD, Weller RO. Prognostic indicators in a range of astrocytic tumours: an immunohistochemical study with Ki-67 and p53 antibodies. J Neurol Neurosurg Psychiatry. 1995;59: 413-419.

[43.] McKeever PE, Ross DA, Strawderman MS, Brunberg JA, Greenberg HS, Junck L. A comparison of the predictive power for survival in gtiomas provided by MIB-1, bromodeoxyuridine and proliferating cell nuclear antigen with histopathologic and clinical parameters. J Neuropathol Exp Neurol. 1997;56:798-805.

[44.] McKeever PE, Strawderman MS, Yamini B, Mikhail AA, Blaivas M. MIB-1 proliferation index predicts survival among patients with grade II astrocytoma. J Neuropathol Exp Neurol. 1998;57:931-936.

[45.] Hsu DW, Louis DN, Efird JT, Hedley-Whyte ET. Use of MIB-1 (Ki-67) immunonoreactivity in differentiating grade II and grade III gliomas. J Neuropathol Exp Neurol. 1997;56:857-876.

[46.] Torp SH. Proliferative activity in human glioblastomas: evaluation of different Ki67 equivalent antibodies. Mol Pathol. 1997;50:198-200.

[47.] Sehgal A. Molecular changes during the genesis of human gtiomas. Semin Surg Oncol. 1998;14:3-12.

[48.] Schuurman PR, Troost D, Verbeeten B Jr, Bosch DA. 5-year survival and clinical prognostic factors in progressive supratentorial diffuse "low-grade" astrocytoma: a retrospective analysis of 46 cases. Acta Neurochir (Wien). 1997;139: 2-7.

[49.] Sullivan KF. Structure and utilization of tubulin isotypes. Annu Rev Cell Biol. 1988;4:687-716.

[50.] Khan IA, Luduena RF. Phosphorylation of [Beta]-tubulin. Biochemistry, 1996; 35:3704-3711.

[51.] Luduena RF. Multiple forms of tubulin: different gene products and covalent modifications. Int Rev Cytol. 1998;178:207-275.

[52.] Fanarraga ML, Avila J, Zabala JC. Expression of unphosphorylated class III [Beta]-tubulin isotype in neuroepithelial cells demonstrates neuroblast commitment and differentiation. Eur J Neurosci. 1999;11:517-527.

[53.] Lopes MBS, Frankfurter A, Zientek GM, Herman MM. The presence of neuron-associated microtubule proteins in the human U-251 MG cell line. A comparative immunoblot and immunohistochemical study. Mol Chem Neuropathol. 1992;17:273-287.

[54.] Zientek GM, Herman MM, Katsetos CD, Frankfurter A. Absence of neuron-associated microtubule proteins in the rat C-6 glioma cell line. A comparative immunoblot and immunohistochemical study. Neuropathol Appl Neurobiol. 1993;19:346-349.

[55.] Ranganathan S, Dexter DW, Benetatos CA, Chapman AE, Tew KD, Hudes GR. Increase of [Beta]III- and Beta]IVa-tubulin isotypes in human prostate carcinoma cells as a result of estramustine resistance. Cancer Res. 1996;56:2584-2589.

[56.] Ranganathan S, Dexter DW, Benetatos CA, Hudes GR. Cloning and sequencing of human Beta]III-tubulin cDNA: induction of [Beta]III isotype in human prostate carcinoma cells by acute exposure to antimicrotubule agents. Biochim Biophys Acta. 1998;1395:237-245.

[57.] Ranganathan S, Benetatos CA, Colarusso PJ, Dexter DW, Hudes GR. Altered [Beta]-tubulin isotype expression in paclitaxel-resistant human prostate carcinoma cells. Br J Cancer. 1998;77:562-566.

[58.] Kavallaris M, Kuo DYS, Burkhart CA, et al. Taxol-resistant epithelial ovarian tumors are associated with altered expression of specific [Beta]-tubulin isotypes. J Clin Invest. 1997;100:1282-1293.

[59.] Kavallaris M, Burkhart CA, Horwitz SB. Antisense oligonucleotides to class III [Beta]-tubulin sensitize drug-resistant cells to Taxol. Br J Cancer. 1999;80:1020-1025.

[60.] Jordan MA, Toso RJ, Thower D, Wilson L. Mechanism of mitotic block and inhibition of cell proliferation by taxol at low concentrations. Proc Natl Acad Sci U S A. 1993;90:9552-9556.

[61.] Lu Q, Luduena RF. Removal of [Beta]III isotype enhances taxol-induced microtubule assembly. Cell Struct Funct. 1993;18:173-182.

[62.] Derry WB, Wilson S, Khan IA, Luduena RF, Jordan MA. Taxol differentially modulates the dynamics of microtubules assembled from unfractionated and purified [Beta]-tubulin isotype. Biochemistry. 1997;36:3554-3562.

[63.] Giannakakou P, Gussio R, Nogales E, et al. A common pharmacophore for epithilone and taxanes: molecular basis for drug resistance conferred by tubulin mutations in human cancer cells. Proc Natl Acad Sci U S A. 2000;97:2904-2909.

[64.] Bodey B, Cosgrove M, Gonzalez-Gomez I, Siegel SE, Martin SE, Gilles FH. Co-expression of four intermediate filament subclasses in childhood glial neoplasms. Mod Pathol. 1991;4:742-749.

[65.] Franke FE, Schachenmayr W, Osborn M, Altmannsberger M. Unexpected immunoreactivities of intermediate filament antibodies in human brain and brain tumors. Am J Pathol. 1991;139:67-79.

[66.] Tlhyama T, Lee VM, Trojanowski JQ. Co-expression of low molecular weight neurofilament protein and glial fibrillary acidic protein in established human glioma cell lines. Am J Pathol. 1993;142:883-892.

[67.] Miller DC, Koslow M, Budzilovich GN, Burstein DE. Synaptophysin: a sensitive and specific marker for ganglion cells in central nervous system neoplasms. Hum Pathol. 1990;21:271-276.

[68.] Wolf HK, Buslei R, Blumcke I, Wiestler OD, Pietsch T. Neural antigens in oligodendrogliomas and dysembryoplastic neuroepithelial tumors. Acta Neuropathol (Bed). 1997;94:436-443.

[69.] Dehghani F, Schachenmayr W, Laun A, Korf HW. Prognostic implication of histopathological, immunohistochemical and clinical features of oligodendrogliomas: a study of 89 cases. Acta Neuropathol (Berl). 1998;95:493-504.

[70.] Wharton SB, Chan KK, Hamilton FA, Anderson JR. Expression of neuronal markers in oligodendrogtiomas: an immunohistochemical study. Neuropathol Appl Neurobiol. 1998;24:302-308.

[71.] Powell SZ, Yachnis AT, Rorke LB, Rojiani AM, Eskin TA. Divergent differentiation in pleomorphic xanthoastrocytoma. Evidence for a neuronal element and possible relationship to ganglion cell tumors. Am J Surg Pathol. 1996;20:80-85.

[72.] Vajtai I, Varga Z, Aguzzi A. Pleomorphic xanthoastrocytoma with gangliogliomatous component. Pathol Res Pract. 1997;193:617-621.

[73.] Lopes MB, Altermatt HJ, Scheithauer BW, Shepherd CW, VandenBerg SR. Immunohistochemical characterization of subependymal giant cell astrocytomas. Acta Neuropathol (Berl). 1996;91:368-375.

[74.] Yeung RS, Katsetos CD, Klein-Szanto A. Cerebral astrocytic hamartomas in the Eker rat: a model of tuberous sclerosis. Am J PathoL 1997;151:1477-1486.

[75.] Schiffer D, Cavalla P, Chio A, et al. Tumor cell proliferation and apoptosis in medulloblastoma. Acta Neuropathol (Berl). 1994;87:362-370.

[76.] Iijima M, Nakazato Y. Pale islands in medulloblastoma consist of differentiated cells with low growth potential. Pathol Int. 1997;47:25-30.

[77.] Schiffer D, Bortolotto S, Bosone I, et al. Cell-cycle inhibitor p27/Kip-1 expression in non-astrocytic and non-oligodendrocytic human nervous system tumors. Neurosci Lett. 1999;264:29-32.

[78.] Hirose T, Scheithauer BW, Lopes MB, Gerber HA, Altermatt HJ, Vanden Berg SR. Ganglioglioma: an ultrastructural and immunohistochemical study. Cancer. 1997;79:989-1003.

Accepted for publication January 2, 2001.

From the Departments of Pediatrics (Drs Katsetos, Legido, and Parikh) and Pathology and Laboratory Medicine (Drs Katsetos and de Chadarevian), St Christopher's Hospital for Children, and Department of Pediatrics, MCP Hahnemann University School of Medicine, Philadelphia, Pa (Drs Katsetos, Legido, and de Chadarevian); Center for Neurovirology and Cancer Biology, Temple University, College of Science and Technology, Philadelphia, Pa (Drs Del Valle, Assimakopoulou, and Khalili); Department of Histopathology and Morbid Anatomy, Queen Mary, University of London, London, England (Dr Geddes); Departments of Anatomy (Drs Assimakopoulou and Varakis) and Neurosurgery (Dr Maraziotis), University of Patras School of Medicine, Patras, Greece; Departments of Pathology (Dr Boyd) and Biology (Ors Spano and Frankfurter), University of Virginia, Charlottesville, Va; Department of Pathology, Microbiology, and Immunology, Philadelphia College of Osteopathic Medicine, Philadelphia, Pa (Dr Balin); Department of Biochemistry, Hellenic Pasteur Institute, Athens, Greece (Dr Matsas); and Neuropathology Section, Clinical Brain Disorders Branch, National Institute of Mental Health, IRP, NIH, Bethesda, Md (Dr Herman).

This work is part of the PhD thesis of Dr Katsetos at The Gade Institute, Department of Pathology, University of Bergen, Bergen, Norway (Preceptor: Prof Sverre J. Mork).

Reprints: Christos D. Katsetos, MD, MRCPath, Section of Neurology/ Research Laboratories, St Christopher's Hospital for Children, Erie Avenue at Front Street, Philadelphia, PA 19134 (e-mail: Christos. [e-mail]).