Journal of Neuropathology and Experimental Neurology
Copyright q 2002 by the American Association of Neuropathologists
Vol. 61, No. 4
April, 2002
pp. 329 338
Molecular Markers that Identify Human Astrocytomas and Oligodendrogliomas
BRIAN POPKO, PHD, DENNIS K. PEARL, PHD, DIANE M. WALKER, BS, THEODORE C. COMAS, MD, PHD,
KRISTINE D. BAERWALD, PHD, PETER C. BURGER, MD, BERND W. SCHEITHAUER, MD, AND
ALLAN J. YATES, MD, PHD
Abstract. The classification of human gliomas is currently based solely on neuropathological criteria. Prognostic and therapeutic parameters are dependent upon whether the tumors are deemed to be of astrocytic or oligodendroglial in origin. We
sought to identify molecular reagents that might provide a more objective parameter to assist in the classification of these
tumors. In order to identify mRNA transcripts for genes normally transcribed exclusively by oligodendrocytes, Northern blot
analysis was carried out on RNA samples from 138 human gliomas. Transcripts encoding the myelin basic protein (MBP)
were found in an equally high percentage of tumors that by neuropathological criteria were either astrocytic or oligodendroglial.
In contrast, proteolipid protein (PLP) and cyclic nucleotide phosphodiesterase (CNP) mRNA molecules were found significantly more often in oligodendrogliomas than in astrocytomas. The strongest association with histological typing was found
with the transcript for the myelin galactolipid biosynthetic enzyme UDP-galactose: ceramide galactosytransferase (CGT),
which was about twice as frequently detected in tumors of oligodendroglial type. Results of glycolipid analyses were previously reported on a subset of the tumors studied herein. Statistical analyses of both molecular and biochemical data on this
subset of astrocytomas, oligoastrocytomas, and oligodendrogliomas were performed to determine if a panel of markers could
be used to separate astrocytic and oligodendroglial tumors. The presence of asialo GM1 (GA1) and the absence of paragloboside occurred most frequently in oligodendrogliomas. Ceramide monohexoside (CMH) levels correlated highly with the
expression of mRNA for 4 myelin proteins: CGT, MBP, CNP, and PLP. The best combination of 2 markers of oligodendroglial
tumors was CGT and GA1; the best combination of 3 markers was the presence of CGT, GA1, and the absence of paragloboside. We conclude that this combination of markers could be useful in distinguishing between astrocytic and oligodendroglial
tumors.
Key Words:
Astrocytoma; Ganglioside; Gene expression; Glioma; Glycolipid; Myelin; Oligodendroglioma.
INTRODUCTION
Astrocytes and oligodendrocytes are the predominant
glia of the central nervous system (CNS) (1). Astrocytes
are believed to play a key role in buffering the extracellular milieu of the mature CNS through the expression
of specific receptors and the formation of a network-like
syncitium (2). Oligodendrocytes are the myelinating cells
of the CNS, ensheathing axons with a specialized, multilayered, membranous wrapping that facilitates saltatory
neuronal conduction (3). The differentiated functions of
these specialized cells are achieved through the expression of a number of cell-specific proteins.
In 1926 Bailey and Cushing (4) described a group of
tumors that they called oligodendroglioma based on their
cytologic appearance. Although this designation is now
widely used, the ‘‘cell of origin’’ of these tumors has not
From the University of North Carolina Neuroscience Center (BP,
DMW, KDB), University of North Carolina, Chapel Hill, North Carolina; Departments of Pathology (TCC, AJY) and Statistics (DKP), The
Ohio State University, Columbus, Ohio; Department of Pathology
(PCB), Johns Hopkins University, Baltimore, Maryland; Department of
Pathology, Mayo Clinic (BWS), Rochester, Minnesota.
Correspondence to: Allan J. Yates, MD, PhD, Division of Neuropathology, Room 4166 Graves Hall, 333 West 10th Avenue, Columbus,
Ohio 43210.
Supported by grants from NCI: UO1-CA50910; UO1-CA64974; UOCA64974; UO1-CA50905; UO1-CA64928; PO1 CA085799; and the
NCI Cooperative Human Tissue Network.
been established. Indeed, since there is at present no practical, definitive marker of neoplastic oligodendrocytes,
the general histological appearances of such tumors places them in a morphological continuum that includes other
glial as well as neuronal tumors. The occurrence of mixed
oligodendroglial and astrocytic tumors further compounds problems in diagnosis. Lastly, since an intimate
mix of neoplastic, reactive, and residual normal cells
comprises the bulk of these infiltrative tumors, the histological features that distinguish oligodendroglial tumors
from other gliomas are not absolute criteria. Furthermore,
they are subject to considerable interobserver interpretation. For example, Coons et al found that among 4 neuropathologists reviewing histological slides of gliomas,
perfect diagnostic concordance occurred only 54% of the
time (5). Diagnostic markers that distinguish lesions in
this spectrum of diffuse gliomas in a biologically and
clinically meaningful way are clearly needed.
The purpose of the present study was to determine
whether a panel of molecular markers correlates with the
diagnoses of 3 neuropathologists experienced in glioma
diagnosis. When such neuropathology expertise is not
available, the availability of such markers might permit
more accurate classification of these tumors. In addition,
pathologic diagnoses incorporating information regarding
these markers might provide a more reliable basis for
comparing results of multi-institutional clinical trials. The
authors have independently studied 2 classes of molecular markers of potential value in defining astrocytomas
329
330
POPKO ET AL
and oligodendrogliomas. One class consists of mRNAs
that code for specific myelin proteins and is reported
herein for the first time. The other is the glycolipid composition of these tumors, the subject of a previous publication (6). To determine whether a combination of
markers optimally define these tumors, we analyzed all
of the data from these 2 studies on an overlapping group
of cases composed of fibrillary astrocytomas (A), oligoastrocytomas (OA), and oligodendrogliomas (O). No
pilocytic astrocytomas were included. Our results indicate
that specific combinations of these markers are characteristic of astrocytomas and others of tumors consisting
entirely or in part of oligodendroglioma. We conclude
that these molecular markers may help define the diagnosis of oligodendroglioma, and provide an adjunct to
morphological diagnosis.
MATERIALS AND METHODS
Tissue Procurement and Neuropathology Review
All tumors studied were obtained from the files of Johns
Hopkins Medical Center, Mayo Clinic, The Ohio State University Medical Center, the NCI Cooperative Human Tissue Network, and the University of North Carolina at Chapel Hill.
Specimens were obtained at the time of surgery, frozen in liquid
nitrogen, stored at 2708, and shipped on dry ice. Representative
formalin-fixed and paraffin-embedded histologic slides stained
by hematoxylin and eosin were independently reviewed by the
3 neuropathologist authors (PCB, BWS, and AJY). The majority opinion of these reviewers served as the final diagnosis.
Relevant information regarding the patient and all specimens
was entered into the OSU Neuro-Oncology Information System
(7).
Myelin Protein Gene cDNA Probes: Human myelin gene
probes were generated from normal human brain cDNA prepared using the SUPERSCRIPTTM Preamplification System
(GibcoBRL, Gaithersburg, MD) according to the manufacturer’s protocols for random hexamer-primed first strand cDNA
synthesis. Gene-specific primers were synthesized based on
published sequence information (myelin basic protein [MBP],
[8]; proteolipid protein [PLP], [9]; cyclic nucleotide phosphodiesterase [CNP], [10]; ceramide galactosyltransferase [CGT],
[11]). Oligonucleotides (GibcoBRL) used for amplification of
target cDNA by polymerase chain reaction (PCR) were as follows: (MBP forward primer) 59-GCAGCCTACCGGCAATTTCC, (MBP reverse primer) 59-CAGCCCGCATGTCACATACC,
(PLP forward primer) 59-GGAGCGGGTGTGTACTTGTTTGG,
(PLP reverse primer) 59TTCCCATTTCTAGCAGGAACCAGC,
(CNP forward primer) 59-AGGACTTCCTGCCGCTCTACTTCGG, and (CNP reverse primer) 59-GTTGCTGCTCGCTTAACTCCACCC. The PCR reactions underwent 25 cycles of 1 min at
948C, 1 min at 558C, and 1 min at 728C. Amplification products
were ligated into the pCR II vector using the TA Cloning System (Invitrogen, Carlsbad, CA) and radiolabeled using PCR as
described (12).
RNA Isolation and Northern Blot Analysis: RNA was isolated
from the tumor samples using either the guanidinium thiocynate
procedure described by (13) or using the TRIzol reagent
(GibcoBRL) in accord with the company protocol. Northern
J Neuropathol Exp Neurol, Vol 61, April, 2002
blots were generated as previously described (14). Blots were
hybridized using the human myelin gene cDNA probes,
washed, and exposed to autoradiographic film. Expression levels were determined for each tumor in comparison to that of
normal human brain. Comparisons were done by eye after normalization with 18S-labeled RNA blots (15). In addition, densitometry measurements were taken for most blots with the aid
of NIH Image (National Institutes of Health, Bethesda, MD).
Glycolipid Analyses
Neutral glycolipids were extracted and analyzed as previously described (6). Briefly, the process is as follows. Frozen tissues were weighed to obtain fresh weights, lyophilized overnight, and reweighed to obtain the dry weight. Dried specimens
were then rehydrated with water (20% of fresh weight) and
extracted with 20 volumes of chloroform/methanol (1:2). The
tissue pellet was then re-extracted with 20 volumes of chloroform/methanol (1:2) and the lipid extracts were combined. Aliquots of this were weighed on a Cahn electronic microbalance;
the nonlipid residue was desiccated and weighed. Total lipid
was separated into neutral and acidic fractions using DEAESephadex column chromatography (16). Neutral lipids were
acetylated, separated on a Florisil column, and deacetylated by
the method of Saito and Hakomori (17). The neutral glycolipid
fraction was desalted by dissolving it in chloroform, followed
by centrifugation. Total neutral glycolipid was quantitated on
the basis of neutral hexose by the orcinol technique (18). Neutral glycolipids were separated using high performance thin layer chromatography (HPTLC), detected with diphenylamine
spray, and quantitated using scanning densitometry (19). Paragloboside and asialoGM1 (GA1) were quantitated by immunoHPTLC using F1H11 (Dainabot Co., Tokyo, Japan) and antiGA1 antiserum (gift of Dr. RK Yu), respectively (19).
Statistical Analyses
Statistical inferences comparing histologic diagnoses of cell
type used an ordinal logistic regression model controlling for
tumor grade and/or for age (treated as a continuous variable in
this set of adult patients). An exploratory survival analysis examined the potential value of mRNA transcripts as prognostic
markers using the Cox proportional hazards model where the
population studied and the covariates included in the model are
specified in the presentation. The analysis of % ceramide monohexoside (CMH) levels versus the levels of each mRNA used
a regression model treating semi-quantitative protein levels as
quantitative explanatory variables (similar results were obtained
using a nonparametric alternative; Jonckheere’s test). Two-sided
p values were used for all of the above analyses. Finally, we
examined residual plots and measures of model fit to ensure
that the data reasonably matched the assumptions underlying
each analysis.
RESULTS
Correlation of Transcript Expression with Diagnosis
and Grade
To evaluate the level of myelin protein gene transcripts
in human tumors of glial type, total RNA was isolated
from each tumor and examined by Northern blot analysis.
MOLECULAR MARKERS FOR HUMAN GLIOMAS
Each blot contained RNA from normal human brain as a
positive control and from liver as a negative control.
Moreover, each blot was stripped and rehybridized with
an end-labeled oligonucleotide probe directed toward the
18S ribosomal RNA to control for loading variability.
Tumor samples that contained detectable expression of
the myelin transcripts (at least 25% of the level detected
in normal human brain) were scored as positive. An example of such analyses is shown in Figure 1.
The 138 tumors studied included 16 anaplastic astrocytomas, 64 glioblastoma multiforme, 7 oligoastrocytomas, 14 anaplastic oligoastrocytomas, 17 oligodendrogliomas, and 20 anaplastic oligodendrogliomas. MBP and
PLP were measured in 137 cases, CNP in 102, and CGT
in 120 cases. Of the 99 cases in which all 4 transcripts
were examined none of the 4 were present in 22 cases,
1 of the 4 in 12 cases, 2 of the 4 in 11, and all 4 transcripts were present in 31 tissue specimens.
The MBP transcript was detected in 63% of the gliomas examined (Table 1). However, based on histopathological diagnosis or grade there was no difference in the
portion of tumors that contained MBP mRNA.
The PLP transcript was found in 66% of cases studied.
Although it was more frequently present in oligodendroglial tumors, this was not statistically significant when
controlling for age and grade.
The CNP transcript was present in about 61% percent
of the gliomas studied. It was present in a significantly
higher proportion of tumors classified as oligodendroglial
than astrocytic, even when controlling for grade (p ;
0.08) and both age and grade (p ; 0.04).
CGT mRNA was detected in approximately twice the
portion of oligodendrogliomas as in astrocytomas, and
the difference was highly significant even controlling for
grade and age together (p ; 0.007), as well as grade
alone (p ; 0.004). Focusing just on grade 3 tumors, we
found that 11 of 13 (85%) oligodendrogliomas and 5 of
6 (83%) oligoastrocytomas were CGT-positive compared
to only 3 of 10 (30%) anaplastic astrocytomas (p ; 0.01).
Correlation of Transcript Expression with Survival
Although this study was designed to examine issues of
cell type, clinical follow-up information was available for
114 cases, thus allowing the performance of exploratory
survival analyses. This included 69 deaths and 45 patients still alive after a median of 3.2 yr postoperative
follow-up. To judge the consistency of our results, survival analyses were also performed separately for the
subset of 64 grade 4 tumors with 49 deaths, for the 61
astrocytic tumors with 46 deaths, and for the combined
group of 53 oligoastrocytomas and oligodendrogliomas
with 23 deaths (Table 2). Analyses for CNP and CGT
involved somewhat fewer cases.
In a proportional hazards model controlling for age,
grade, and cell type, tumors with MBP, CNP, or CGT
331
show significantly greater risks than tumors without these
transcripts (Table 2). The degree of statistical significance
was less for the latter 2, partly due to the smaller sample
sizes. Expression of the transcript for PLP was not associated with survival. Figure 2 shows these results for
the presence and absence of MBP as a Kaplan-Meier plot
for the 32 grade 4 tumors of patients over 60 yr (p ;
0.037 by logrank test).
Combinations of mRNA Expression and Glycolipids
as Markers
We statistically analyzed an overlap subset of 74 gliomas that were part of the series examined for myelin
protein transcript expression in the current study and for
glycolipid composition as part of a previous report (6).
This overlap group consisted of 33 fibrillary astrocytomas (1 grade 2; 6 grade 3; 26 grade 4), 25 oligodendrogliomas (12 grade 2; 9 grade 3; 4 grade 4) and 16 oligoastrocytomas (2 grade 2; 9 grade 3; 5 grade 4).
In the study of glycolipids it was found that (a) the
presence of GA1 and high levels (over 50% of the total
neutral glycolipid) of CMH were present more frequently
in oligodendrogliomas than astrocytomas, and (b) paragloboside was present in most astrocytomas, but in only
a few oligodendrogliomas. As in the complete group of
138 tumors, in this subset of 74 gliomas messages for
the 3 myelin proteins CGT, PLP, and CNP, were all found
more often in oligodendroglial than astrocytic tumors,
while MBP was not. Of these oligodendroglial markers,
CGT had the strongest positive association with tumors
having an oligodendroglioma component as compared to
pure astrocytomas (Table 3).
Galactosylcerebroside is a product of the biochemical
reaction catalyzed by CGT and is by far the most abundant, but not only, ceramide monohexoside (CMH) in
most human gliomas (19). However, because the chromatographic system we used does not separate galactosylcerebroside from glucosylceramide, we use the term
CMH in our results when referring to these compounds.
Both galactosylcerebroside and CGT are markers of oligodendroglia and myelin, as are MBP, CNP, and PLP.
Thus it is not surprising that the proportion of total neutral glycolipids accounted for by CMH correlated with
the amount of message for these 4 myelin protein markers (Table 4). The proportion of CMH also correlated
with the presence of GA1, but was inversely related to
the presence of paragloboside. The presence of PLP, CGT,
CNP, and MBP all correlated strongly with each other
(data not shown).
The 3 constituents that individually best differentiated
between oligodendrogliomas and astrocytomas were
CGT, GA1, and paragloboside (Table 3). We also sought
combinations of markers that best distinguished between
these 2 tumors. The best combination of 2 markers for
oligodendrogliomas was CGT and GA1 (Table 5). Of 29
J Neuropathol Exp Neurol, Vol 61, April, 2002
332
POPKO ET AL
Fig. 1. Northern blot analysis of oligodendroglial marker transcripts in human tumors. A: Five mg RNA samples isolated
from human tumor and tissue samples were hybridized sequentially to the indicated radiolabeled cDNA probes. Subsequently the
blot was hybridized to an oligonucleotide probe specific for the 18S ribosomal RNA species to control for loading variability.
Histological diagnosis of the tumors are as follows: lane 1, anaplastic astrocytoma; lane 2, oligodendroglioma; lane 3, glioblastoma
multiforme; lane 4, glioblastoma multiforme; lane 5, pilocytic astrocytoma; lane 6, glioblastoma multiforme; lane 7 normal human
liver and lane 8 normal human brain. For reference, lanes 1, 3, 4, and 5 are positive for MBP; lanes 1, 2, 3, and 5 are positive
for PLP; and lanes 1, 2, 3, 4, and 5 are positive for CNP. B: Five mg RNA samples isolated from human tumor and tissue samples
were hybridized to the human CGT cDNA probe and then to the 18S ribosomal probe. Histological diagnosis of the tumors are
as follows: lane 1, anaplastic oligodendroglioma; lane 2, anaplastic oligodendroglioma; lane 3, glioblastoma multiforme; lane 4,
anaplastic mixed oligodendroglioma-astrocytoma; lane 5, astrocytoma; lane 6, anaplastic mixed oligodendroglioma-astrocytoma;
lane 7, glioblastoma multiforme; lane 8, oligodendroglioma; lane 9, ependymoma; lane 10, anaplastic oligodendroglioma; lane
11, normal human brain; and lane 12 normal human liver. For reference, lanes 1, 2, 4, 6, 8, and 10 are positive for the CGT
transcript.
J Neuropathol Exp Neurol, Vol 61, April, 2002
333
MOLECULAR MARKERS FOR HUMAN GLIOMAS
TABLE 1
Association between the Presence of mRNA Transcripts and Histologic Diagnosis
Diagnosis
MBP
Anaplastic Astrocytoma
Glioblastoma
Oligoastrocytoma
Anaplastic Oligoastrocytoma
Oligodendroglioma
Anaplastic Oligodendroglioma
All cases
p-valuea,b
16
64
7
14
17
19
137
0.27
PLP
16
64
7
13
17
20
137
0.44
69%
61%
86%
64%
59%
58%
63%
0.31
CNP
63%
59%
86%
62%
77%
80%
66%
0.37
8
45
4
12
14
19
102
0.04
CGT
63%
47%
75%
50%
79%
84%
61%
0.08
10
62
5
10
15
18
120
0.007
30%
37%
40%
70%
67%
78%
49%
0.004
Values represent the number of cases studied and the percentage that showed presence of the mRNA transcript for the designated
protein. P values are the results of ordinal logistic regression analyses controlling for aage and grade; bgrade. However, age was
not a significant factor in determining cell type after controlling for grade, so analyses with and without age in the model gave
similar results.
TABLE 2
Associations among the Presence of mRNA Transcripts and Patient Survival
Population studied;
variables in model
All cases; age, grade, cell type, transcript
Grade 4; age, cell type, transcript
O and OA; age, grade, transcript
A: age, grade, transcript
MBP
n 5 113
0.017
1.89
(1.12;
0.010
2.26
(1.21;
0.083
2.24
(0.90;
0.11
1.69
(0.89;
3.28)
4.38)
6.38)
3.32)
PLP
n 5 113
0.27
1.35
(0.80;
0.29
1.40
(0.75;
0.54
1.40
(0.50;
0.39
1.32
(0.71;
2.32)
2.66)
5.03)
2.53)
CNP
n 5 92
0.026
2.01
(1.08;
0.014
2.53
(1.21;
0.35
1.66
(0.60;
0.024
2.53
(1.13;
3.79)
5.37)
5.95)
5.69)
CGT
n 5 104
0.086
1.58
(0.94;
0.19
1.52
(0.80;
0.074
2.58
(0.92;
0.36
1.36
(0.69;
2.66)
2.82)
9.18)
2.58)
Each cell presents the 2-sided p-value from a Cox model, the estimated risk ratio for the transcript, and a 95% confidence
interval for the risk ratio (in parentheses). Risk ratio .1 indicates that the presence of the mRNA transcript is associated with a
shorter survival. O 5 oligodendroglioma; OA 5 oligoastrocytoma; A 5 astrocytoma.
astrocytomas, only 1 contained both of these substances,
whereas all of the 30 tumors with an oligodendroglioma
component had at least 1 of these markers. The data were
analyzed by an ordinal logistic regression, the results of
which demonstrated a highly significant (p ; 0.002) correlation of the combination of CGT and GA1 with an
oligodendroglioma component, even after controlling for
age and grade. The combination of CGT and absence of
paragloboside was almost as good at differentiating between astrocytomas and oligodendroglioma-containing
tumors (p ; 0.005). Adding the absence of paragloboside
as an oligodendroglioma marker provided significant (p
; 0.001) additional diagnostic information over the combination of CGT and GA1 (Table 6). In contrast, individually adding CMH, CNP, or PLP to this 3-way combination provided no additional diagnostic information.
Although MBP showed no relationship to cell type alone,
surprisingly, it was a negative marker of an oligodendroglioma component when viewed in combination with
CGT, GA1 and absence of paragloboside (data not
shown).
DISCUSSION
The oligodendrocyte was first identified by Robertson
in 1900 (20) and so named in 1921 by Hortega (21).
However, it was not until the 1960s that evidence was
obtained from electron microscopic studies that central
nervous system myelin is both produced by and a part of
the oligodendrocyte (22). Nonetheless, on the basis of
metallic impregnation studies, Bailey and his coworkers
had even earlier concluded that there exists a group of
tumors in which neoplastic oligodendrocytes predominate
(4, 23). Interestingly, they and Ravens et al (24) found
that such neoplasms contain many tumor cells that appeared to be of astrocytic origin or intermediate between
these 2 glia in appearance. Although this should not be
surprising, as studies on the cytogenesis of astrocytes and
J Neuropathol Exp Neurol, Vol 61, April, 2002
334
POPKO ET AL
Fig. 2. Kaplan-Meier survival plots for patients over 60 yr
old with grade 4 tumors with and without the presence of the
MBP transcript. The lower curve consists of 16 glioblastomas,
2 anaplastic oligoastrocytomas, and 1 anaplastic oligodendroglioma. The upper curve consists of 11 glioblastomas, 1 anaplastic oligoastrocytoma, and 1 anaplastic oligodendroglioma.
oligodendroglia have demonstrated that their development is closely related, it does explain difficulties encountered in histological diagnosis and in classification
of the diffuse gliomas.
Because evidence supporting the oligodendrocyte as
the ‘‘cell of origin’’ of oligodendrogliomas is inconsistent, the cytogenesis of these tumors remains a debated
issue. Based on the study of formalin-fixed, paraffin-embedded tissue, the principle histologic attributes of these
neoplasms are the presence of clear cells with uniform,
round nuclei. It is important to note that these cytologic
characteristics are entirely nonspecific. Indeed, several
clear cell tumors with light microscopic appearances similar to oligodendrogliomas are not at all oligodendroglial
in origin. Examples include clear cell ependymoma (25),
central neurocytoma (26), dysembryoplastic neuroepithelial tumor (27), and clear cell meningioma (28). Ultrastructural studies demonstrating myelin-like spirals in oligodendrogliomas are rare. Although this finding has been
used to support an oligodendroglial cytogenesis, it can
also be interpreted as aberrant myelination by traumatized, non-neoplastic oligodendroglia (29). On the basis
of immunostaining for synaptophysin, Miller claims that
over half of cases diagnosed as oligodendroglioma or oligoastrocytoma by routine histology are neuronal tumors,
ones he terms ‘‘ganglioneurocytoma’’ (30). The electrophysiologic study of Patt et al supports the view that a
proportion of what are called ‘‘oligodendrogliomas’’ possess neuronal properties (31). These authors found that
in some cases both cells in slices of fresh oligodendroglioma tissue as well as cells cultured from surgical specimens of oligodendrogliomas and oligoastrocytomas generate action potentials similar to those of neurons.
J Neuropathol Exp Neurol, Vol 61, April, 2002
Oligodendrogliomas generally have a higher proportion
of 1b gangliosides than do astrocytomas, as do neurons
in comparison with astrocytes (6). It is intriguing, but
unknown, if this represents a neuronal relationship of
these tumors.
In one immunohistochemical study for myelin basic
protein (MBP) in oligodendrogliomas, the investigators
interpreted their results as indicating that all such tumors
express this protein (32). In contrast, Nakagawa et al (33)
found their series of oligodendrogliomas to express neither MBP nor myelin-associated glycoprotein (MAG).
Lastly, the study of Sung et al found 5 of 22 oligodendrogliomas with immunoreactivity for galactosylcerebroside, and only 1 of 24 cases for sulfatide, a sulfated form
of galactosylcerebroside (34). They found only 1 case
that expressed MBP and CNP, and both of these antigens
were weakly expressed. These authors interpreted their
findings in a manner similar to those of Figols et al (32),
attributing positivity to the presence of myelin debris
both within histiocytes and extracellularly. To date, cells
cultured from oligodendrogliomas and oligoastrocytomas
also have not been shown to express oligodendrocytespecific antigens.
These immunohistochemical results are somewhat in
contrast to the present study in which we examined human glial tumors for the presence of myelin protein gene
transcripts using Northern blot analysis. Our results demonstrate that the presence of mRNA molecules for the
PLP, CNP, and CGT genes correlates with a neuropathological classification of oligodendroglioma, whereas the
MBP transcripts were found in an equally high portion
of astrocytic tumors as in oligodendrogliomas. Of the
various markers, the presence of the CGT transcript provided the best indicator of tumor type. In conjunction
with the aforementioned immunohistochemical studies,
our results suggest that the myelin proteins are unstable
in transformed oligodendrocytes, whereas their corresponding mRNAs likely have a longer half-life. This is
consistent with data that we have generated with transgenic mice in which oncogene-induced oligodendrogliomas were shown to express very low levels of the myelin
proteins, yet the myelin protein transcripts were abundant
(35).
MBP, PLP, and MAG are expressed rather late in the
development of oligodendroglia. On the other hand, CGT,
galactosylcerebroside, and CNP are expressed by immature oligodendrocytes. From our analyses of these tumors, we found that the latter 3 substances are more frequently expressed in oligodendrogliomas than
astrocytomas. This suggests that oligodendrogliomas may
arise from cells at an early stage of oligodendrocyte development, and is consistent with the finding that many
oligodendrogliomas contain glial fibrillary acidic protein
(GFAP) (33, 36–39), a protein transiently expressed by
immature oligodendroglia (38, 40), The expression of
335
MOLECULAR MARKERS FOR HUMAN GLIOMAS
TABLE 3
Association of Oligodendroglioma Markers with Diagnosisa
Fibrillary
Astrocytomas
Oligoastrocytomas
Oligodendrogliomas
CGT
Absent
Presentb
26 (81%; 72%)
6 (19%; 20%)
2 (17% 6%)
10 (83%; 33%)
8 (36%; 22%)
14 (64%; 47%)
PLP
Absent
Present
16 (48%; 62%)
17 (52%; 36%)
4 (27%; 15%)
11 (73%; 23%)
6 (24%; 23%)
19 (76%; 40%)
CNP
Absent
Present
19 (59%; 65%)
13 (41%; 32%)
6 (40%; 21%)
9 (60%; 22%)
4 (17%; 14%)
19 (83%; 46%)
MBP
Absent
Present
12 (36%; 44%)
21 (64%; 46%)
4 (25%; 15%)
12 (75%; 26%)
11 (46%; 41%)
13 (54%; 28%)
GA1
Absent
Present
15 (50%; 75%)
15 (50%; 34%)
1 (8%; 5%)
12 (92%; 27%)
4 (19%; 20%)
17 (81%; 39%)
CMHb
Low (,50%)
High (.50%)
15 (46%; 68%)
18 (54%; 35%)
2 (12%; 9%)
14 (88%; 27%)
5 (20%; 23%)
20 (80%; 38%)
Paragloboside
Absent
Present
9 (31% 26%)
20 (69%; 69%)
9 (69%; 27%)
4 (31%; 14%)
16 (76%; 47%)
5 (24%; 17%)
1b Gangliosidesc
Low (,30%)
High (.30%)
29 (88%; 69%)
4 (12%; 13%)
6 (37%; 14%)
10 (63%; 31%)
7 (28%; 17%)
18 (72%; 56%)
Markers
a
The results are only for the overlap set of tumors studied for both mRNA expression and glycolipid composition. Values
ahead of parentheses are number of tumors. The first number in parentheses is the percentage of tumors in the cell; the second
number in parentheses is the percent of tumors in that row.
b
Only 4 of the 26 grade 4 astrocytomas were positive for CGT.
c
Levels of CMH and 1b gangliosides were determined as a percentage of total neutral glycolipid and gangliosides, respectively.
TABLE 5
CGT and GA1 Combined as Oligodendroglioma
Markersa
TABLE 4
Correlation between Percent CMH and mRNA Levels
for Myelin Proteins
Absent
MBP
PLP
CNP
CGT
56.2
n
53.0
n
57.4
n
61.2
n
6
5
6
5
6
5
6
5
6.5
26
7.0
24
6.0
27
5.4
34
Low
66.5
n
70.5
n
72.6
n
67.8
n
6
5
6
5
6
5
6
5
Medium/High p value*
6.0
24
5.6
27
5.3
27
6.2
16
84.9
n
79.7
n
77.9
n
74.5
n
6
5
6
5
6
5
6
5
3.3
20
4.0
19
8.4
13
8.6
14
0.001
0.003
0.03
0.16
Values are percent of total neutral glycolipid comprised by
CMH (means 6 SEM) in cases with absent, low, or medium to
high levels of message for each protein listed.
* p values were evaluated using a regression analysis.
genes coding for other intermediate filament proteins by
cells cultured from human oligodendrogliomas is also
consistent with such an origin (41). Furthermore, this notion of a ‘‘cell of origin’’ is supported by the study of
Number
of oligo
markers
0
1
2b
Total
Fibrillary
astrocytomas
Oligoastrocytomas
Oligodendrogliomas
12 (100%; 41%)
0
0
16 (59%; 55%)
2 (7%; 18%)
9 (33%; 47%)
1 (5%; 3%)
9 (45%; 82%) 10 (50%; 53%)
29 (49%; 100%) 11 (19%; 100%) 19 (32%; 100%)
a
Values ahead of parentheses are number of tumors. The first
number in parenthesis is percent of row; the second number is
percent of column.
b
Only 1 of 24 grade 4 astrocytomas was positive for both
CGT and GA1.
Nishiyama et al (42) who found that the NG2 proteoglycan and the platelet-derived growth factor receptor alpha,
both of which define an oligodendrocyte progenitor cell,
are expressed in human oligodendrogliomas. All of these
J Neuropathol Exp Neurol, Vol 61, April, 2002
336
POPKO ET AL
TABLE 6
CGT, GA1 and Absent Paragloboside Combined as
Oligodendroglioma Markersa
Number
of oligo
markers
0
1
2b
3c
Total
Fibrillary
astrocytomas
Oligoastrocytomas
Oligodendrogliomas
9 (100%; 32%)
0
0
14 (82%; 50%)
0
3 (18%; 16%)
5 (26%; 18%)
6 (32%; 55%)
8 (42%; 42%)
0
5 (38%; 46%)
8 (62%; 42%)
28 (48%; 100%) 11 (19%; 100%) 19 (33%; 100%)
a
Values ahead of parentheses are number of tumors. The first
number in parenthesis is percent of row; the second number is
percent of column. Presence of CGT and GA1, but absence of
paragloboside, are considered markers of oligodendroglioma.
b
Only 3 of 23 grade 4 astrocytomas had 2 of the 3 oligodendroglioma markers.
c
None of the 23 grade 4 astrocytomas had all 3 oligodendroglioma markers.
more recent findings support interpretations of earlier
morphological studies in which cells along a spectrum of
differentiation, ranging from ones very immature to ones
resembling mature oligodendroglia and astrocytes, were
seen in the same tumor (4, 23, 24). The significance of
the expression of genes for myelin constituents in astrocytomas is uncertain, but some of this gene expression
could represent the activity of residual nonneoplastic oligodendrocytes in the tumor tissues. The unlikely possibility that contaminating non-neoplastic tissue accounts
for a significant component of glioma glycolipids has
been thoroughly discussed in a previous publication (6).
Furthermore, such expression has also been seen in continuous cell lines derived from human astrocytomas (43),
so it seems to be a general property of neoplastic cells in
diffuse gliomas.
Although aberrant gene expression by a genetically abnormal tumor cell is an explanation that cannot be ruled
out, the strong concordance among the myelin markers
examined using distinctly different techniques makes this
an unlikely possibility. Instead, on the basis of morphologic, immunohistochemical, biochemical, and molecular
genetic data, we postulate that astrocytomas and oligodendrogliomas are both derived from glial precursor cells
that have the potential for divergent glial differentiation.
Tumors derived from such cells should consist of variable
proportions of tumor cells at different stages of differentiation along multiple lines. This may explain the occurrence of GFAP-positive oligodendrogliomas, as well
as the detection of cerebroside and sulfatide and the expression of myelin specific genes in some astrocytomas.
This explanation is also consistent with the finding that
expression of the transcript for MBP does not correlate
with a diagnosis of oligodendroglioma but is associated
with poorer survival (Tables 1, 2; Fig. 1).
J Neuropathol Exp Neurol, Vol 61, April, 2002
As it is difficult, if not impossible, to identify these
phenotypically different cells at the light microscopic or
ultrastructural level, it seems reasonable to suggest that
tumors should be classified according to combinations of
molecular markers distinctive for these cell types. Genes
encoding the myelin specific proteins discussed above are
good candidates. Some glycosphingolipids, such as A2B5
epitope, GD3, galactosylcerebroside, and sulfatide, have
also been found to be useful markers of specific stages
of gliogenesis. This forms part of the rationale for our
study, which sought to determine whether specific glycolipid profiles correlate with specific glioma types. In
previous studies we found that the presence of a sialylated derivative of paragloboside (69LM1) in astrocytomas correlated with higher histological grade and a poor
prognosis. Therefore, we studied its precursor, paragloboside, to see if it was of diagnostic or prognostic value.
This proved to be the case, as paragloboside was expressed much more frequently in astrocytomas than oligodendrogliomas. Conversely, GA1 was present more
frequently in oligodendrogliomas than in astrocytomas.
Paragloboside and GA1 are glycosphingolipids that are
synthesized along 2 different metabolic pathways, but
gangliosides of both families can occur in the same cells
(44). Thus it is interesting to find that paragloboside and
GA1 are characteristic of phenotypically different tumor
cells.
Using the results from both molecular genetic and glycolipid analytical studies, we sought to determine whether specific combinations of those best correlating with
histological diagnoses had more diagnostic power than
any one did individually. Two-way combinations of CGT
with GA1, paragloboside, or CMH were better discriminants of astrocytomas and oligodendrogliomas than either
alone (Table 5); three-way combinations were better still
(Table 6). The best 3-way combination was the presence
of both CGT and GA1 coupled with absence of paragloboside. Only astrocytomas were completely devoid of
these 3 oligodendroglioma markers; no astrocytoma expressed all 3. For grade 4 astrocytomas, only 4 of 26 had
CGT (Table 3), and only 1 of 24 had both CGT and GA1
(Table 5). Conversely, only tumors with an oligodendroglioma component exhibited all 3 of these markers. Between these 2 extremes the proportions of tumors with
each of the other 2 combinations of markers was well
ordered within diagnostic categories.
Gangliosides are another class of glycolipids that relate
to the phenotypes of some human gliomas. Gangliosides
of the 1b pathway are normal constituents of the adult
mammalian central nervous system, comprising about
30% of the total ganglioside content (45). We found that
low levels of 1b pathway gangliosides (,30% of total
gangliosides) were observed far more often in astrocytomas than in tumors with oligodendroglioma elements
(Table 3). Results of a logistic regression analysis show
MOLECULAR MARKERS FOR HUMAN GLIOMAS
this relationship to be highly significant, even after controlling for patient age (p value ; 0.0004). In our analysis
there is an almost 10-fold increase (95% C.I. 2.8 to 32.3)
in the odds of a tumor being an astrocytoma rather than
an oligodendroglioma when the 1b pathway gangliosides
comprises less than 30% of the total ganglioside content.
Furthermore, we found that the loss of 1b gangliosides
from astrocytomas correlated directly with histological
grade and inversely with patient survival (46, 47). However, a correlation between 1b ganglioside content and
grade of oligodendrogliomas was not statistically significant (34). This finding was recently confirmed in an immunohistochemical study using an antibody directed
against GD1b, a major ganglioside of the 1b pathway
(48).
In summary, we conclude that both individually and
in combination, a group of 3 molecular markers can be
useful in establishing the diagnosis of astrocytoma and
oligodendroglioma. All of these markers are present in a
higher proportion of tumors designated by 3 review neuropathologists as containing oligodendrogliomatous components. Furthermore, combinations of these markers are
better discriminants than any single marker. Lastly, having established that the cell type is astrocytoma, histological grade can be confirmed by biochemically or immunohistochemically determining levels of 1b
gangliosides (48).
REFERENCES
1. Hall ZW. The cells of the nervous system. In: Sunderland MA, ed.
An introduction to molecular neurobiology. Place: Sinauer Association, Inc., 1992:1–29
2. Shao Y, McCarthy KD. Plasticity of astrocytes. Glia 1994;11:147–55
3. Morell P, Quarles RH, Norton WT. Myelin formation, structure, and
biochemistry. In: Siegel GJ, et al, eds. Basic neurochemistry. Molecular, cellular, and medical aspects. New York: Raven Press,
1994:117–43
4. Bailey P, Cushing H. A classification of tumors of the glioma group
on a histogenetic basis with a correlated study of prognosis. Philadelphia, London, Montreal: Lippincott, 1926
5. Coons SW, Johnson PC, Scheithauer BW, Yates AJ, Pearl DK. Improving diagnostic accuracy and interobserver concordance in the
classification and grading of primary gliomas. Cancer 1997;79:
1381–93
6. Yates AJ, Comas TC, Scheithauer BW, Burger PC, Pearl DK. Glycolipid markers of astrocytomas and oligodendrogliomas. J Neuropathol Exp Neurol 1999;58:1250–62
7. Mamrak SA, Yates AJ. An information system to support collaborative brain tumor research. Brain Tumor Pathol 1998;14:131–37
8. Kamholz J, de Ferra F, Puckett C, Lazzarini R. Identification of
three forms of human myelin basic protein by cDNA cloning. Proc
Natl Acad Sci USA 1986;83:4962–66
9. Kronquist KE, Crandall BF, Macklin WB, Campagnoni AT. Expression of myelin proteins in the developing human spinal cord:
Cloning and sequencing of human proteolipid protein cDNA. J
Neurosci Res 1987;18:395–401
10. Kurihara T, Takahashi Y, Nishiyama A, Kumanishi T. cDNA cloning and amino acid sequence of human brain 29,39-cyclic-nucleotide
39-phosphodiesterase. Biochem Biophys Res Commun 1988;152:
837–42
337
11. Kapitonov DE, Yu RK. Cloning, characterization, and expression
of human ceramide galactosyltransferase cDNA. Biochem Biophys
Res Commun 1997;232:449–53
12. Jansen R, Ledley FD. Production of discrete high specific activity
DNA probes using the polymerase chain reaction. Gene Anal Tech
1989;6:79–83
13. Chirgwin JM, Przybyla AE, MacDonald RJ, Rutter WJ. Isolation
of biologically active ribonucleic acid from sources enriched in
ribonuclease. Biochemistry 1979;18:5294–99
14. Popko B, Puckett C, Lai E, et al. Myelin deficient mice: Expression
of myelin basic protein and generation of mice with varying levels
of myelin. Cell 1987;48:713–21
15. Stahl N, Harry J, Popko B. Quantitative analysis of myelin protein
gene expression during development in the rat sciatic nerve. Mol
Brain Res 1990;8:209–11
16. Ando S, Yu RK. Isolation and characterization of a novel trisialoganglioside GT1a from human brain. J Biol Chem 1977;252:
6247–50
17. Saito T, Hakomori S-I. Quantitative isolation of total glycosphingolipids from animal cells. J Lipid Res 1971;12:257–59
18. Neskovic N, Sarlieve L, Nussbaum JL, Kostic D, Mandel P. Quantitative thin-layer chromatography of glycolipids in animal tissues.
Clinica Chimica Acta 1972;38:147–53
19. Singh LPK, Pearl DK, Franklin TK, et al. Neutral glycolipid composition of primary human brain tumors. Mol Chem Neuropathol
1994;21:241–57
20. Robertson WF. A microscopic demonstration of the normal and
pathological histology of mesoglial cells. Journal of Mental Science
1900;46:724
21. Hortega DR. La glia de escasas (oligodendroglia). Boletin de la
Real Soc espanola de Hist Nat 1921;21:63–92
22. Bunge RP. Glial cells and the central myelin sheath. Physiological
Reviews 1968;48:197–251
23. Bailey P, Bucy PC. Oligodendrogliomas of the brain. Journal of
Pathology and Bacteriology 1929;32:735–51
24. Ravens JR, Adamkiewicz LL, Groff R. Cytology and celluar pathology of the oligodendrogliomas of the brain. J Neuropathol Exp
Neurology 1955;14:142–84
25. Kawano N, Yada K, Yagishita S. Clear cell ependymoma. A histological variant with diagnostic implications. Virchows Archiv A
Pathol Anat 1989;415:467–72
26. Giangaspero F, Cenacchi G, Losi L, Cerasoli S, Bisceglia M, Burger
PC. Extraventricular neoplasms with neurocytoma features. A clinicopathological study of 11 cases. Am J Surg Pathol 1997;21:206–12
27. Daumas-Duport C. Dysembryoplastic neuroepithelial tumours.
Brain Pathol 1993;3:283–95
28. Zorludemir S, Scheithauer BW, Hirose T, Van Houten C, Miller G,
Meyer FB. Clear cell meningioma. A clinicopathological study of
a potentially aggressive variant of meningioma. Am J Surg Pathol
1995;19:493–505
29. Robertson DM, Vogel FS. Concentric lamination of glial processes
in oligodendrogliomas. J Cell Biology 1962;15:313–34
30. Miller DC. New perspectives on neuronal neoplasms. Brain Pathol
1997;7:1139–40
31. Patt S, Labrakakis C, Bernstein M, et al. Neuron-like physiological
properties of cells from human oligodendroglial tumors. Neuroscience 1996;71:601–11
32. Figols J, Iglesias-Rozas JR, Kazner E. Myelin basic protein (MBP)
in human gliomas: A study of twenty-five cases. Clin Neuropathol
1985;4:116–20
33. Nakagawa Y, Perentes E, Rubinstein LJ. Immunohistochemical
characterization of oligodendrogliomas: An analysis of multiple
markers. Acta Neuropathol (Berl) 1986;72:15–22
34. Sung C-C, Collins R, Li J, et al. Glycolipids and myelin proteins
in human oligodendrogliomas. Glycoconj J 1996;13:433–43
J Neuropathol Exp Neurol, Vol 61, April, 2002
338
POPKO ET AL
35. Hayes C, Kelly D, Murayama S, Komiyama A, Suzuki K, Popko
B. Expression of the neu oncogene under transcriptional control of
the myelin basic protein gene in transgenic mice: Generation of
transformed glial cells. J Neurosci 1992;31:175–87
36. De Armond SJ, Eng LF, Rubinstein LJ. The application of glial
fibrillary acidic (GFA) protein immunohistochemistry in neurooncology. Pathol Res Pract 1980;168:374–94
37. Herpers MJHM, Budka H. Glial Fibrillary acidic protein (GFAP)
in oligodendroglial tumors: Gliofibrillary oligodendroglioma and
transitional oligoastrocytoma as subtypes of oligodendroglioma.
Acta Neuropathol (Berl) 1984;64:265–72
38. Jagadha V, Halliday WC, Becker LE. Glial fibrillary acidic protein
(GFAP) in oligodendrogliomas: A reflection of transient GFAP expression by immature oligodendroglia. Can J Neurol Sci 1986;13:
307–11
39. Sarkar C, Roy S, Tandon PN. Oligodendroglial tumors. An immunohistochemical and electron microscopic study. Cancer 1988;
61:1862–66
40. Choi BH, Kim RC. Expression of glial fibrillary acidic protein by
immature oligodendroglia and its implications. J Immunol 1985;8:
215–35
41. Kashima T, Vinters HV, Campagnoni AT. Unexpected expression
of intermediate filament protein genes in human oligodendroglioma
cell lines. J Neuropathol Exp Neurol 1995;54:23–31
J Neuropathol Exp Neurol, Vol 61, April, 2002
42. Nishiyama A, Chang A, Trapp BD. NG21 glial cells: A novel glial
cell population in the adult brain. J Neuropathol Exp Neurol 1999;
58:1113–24
43. Kashima T, Tiu SN, Merrill JE, Vinters HV, Dawson G, Campagnoni AT. Expression of oligodendrocyte-associated genes in cell
lines derived from human gliomas and neuroblastomas. Cancer Res
1993;53:170–75
44. Chou DKH, Flores S, Jungalwala FB. Identification of disialosyl
paragloboside and O-acetyldisialosyl paragloboside in cerebellum
and embryonic cerebrum. Journal of Neurochemistry 1990;54:
1598–607
45. Ando S. Gangliosides in the nervous system. Neurochem Int 1983;
5:507–37
46. Sung CC, Pearl DK, Coons SW, Scheithauer BW, Johnson PC, Yates AJ. Gangliosides as diagnostic markers of human astrocytomas
and primitive neuroectodermal tumors. Cancer 1994;74:3010–22
47. Sung CC, Pearl DK, Coons SW, et al. Correlation of ganglioside
patterns of primary brain tumors with survival. Cancer 1995;75:
851–59
48. Comas TC, Tai T, Kimmel D, et al. Immunohistochemical staining
for ganglioside GD1b as a diagnostic and prognostic marker for
primary human brain tumors. Neuro-Oncology 1999;1:261–67
Received July 31, 2001
Revision received December 4, 2001
Accepted December 13, 2001