Michael Wegner, Institut für Biochemie, Universität Erlangen, Fahrstrasse 17, D-91054 Erlangen, Germany. Tel: +49 9131 85 24620; Fax: +49 9131 85 22484; E-mail: firstname.lastname@example.org
Members of group E and group D of the Sox gene family function as important transcriptional regulators of glial development in the central nervous system. Here, we have examined Sox gene expression in 60 human primary gliomas. Transcripts from each of the six group E and group D genes were expressed in gliomas of various types and malignancy grades, but with significant differences. SOX5, SOX9 and SOX10 were generally expressed at levels similar to or below those in adult brain tissue. In contrast, many oligodendrogliomas exhibited upregulation of SOX6, SOX8 and SOX13. Furthermore, loss of heterozygosity on chromosomal arms 1p and 19q was associated with significantly higher SOX8 mRNA levels. Low-grade astrocytomas, but not glioblastomas, also showed elevated SOX8 transcript levels. Taken together, the expression pattern of Sox genes in gliomas is heterogeneous and overall compatible with the less differentiated state of glioma cells as compared with their normal adult counterparts. Despite their restricted expression in astrocytes and oligodendrocytes during normal development, none of the Sox genes was selectively expressed in tumours of the oligodendroglial or astrocytic lineage. This is compatible with an origin of gliomas from neuroepithelial stem or precursor cells.
Gliomas represent the main type of primary brain tumour and are often associated with fatal outcome because of their highly invasive growth and their frequent resistance to therapy. Gliomas are histologically classified and graded from grade I to IV according to the criteria of the World Health Organization (WHO) classification of tumours of the nervous system . Histological classification and grading are of paramount importance with respect to postoperative treatment and outcome of glioma patients. For example, oligodendroglial tumours more frequently respond to radio- and chemotherapy than astrocytic tumours, and are associated with significantly longer survival [2–4]. Furthermore, patients with low-grade glioma (WHO grades I and II) have a much better prognosis than patients with anaplastic glioma (WHO grade III) or glioblastoma multiforme (WHO grade IV) . The WHO classification of gliomas is primarily based on morphological similarities between the tumour cells and the normal glial cells of the central nervous system (CNS), with tumour cells in oligodendrogliomas sharing some common features with oligodendrocytes, while neoplastic cells in astrocytomas share certain morphological and immunohistochemical characteristics with normal astrocytes. This resemblance has long been taken as indicative of the tumour's origin. However, the recent identification of CD133-positive cells with stem cell-like characteristics in gliomas, as well as studies on experimental gliomas in mice, has fostered the hypothesis that gliomas may arise from neuroectodermal stem or progenitor cells . So far, the precise roles of stem cells in gliomas are far from being completely understood, but likely to be complex.
Attempts are also under way to identify molecular markers for gliomas that may not only facilitate classification and grading, but would also improve our understanding of the origins and molecular pathogenesis of the diverse types of astrocytic and oligodendroglial gliomas. In this respect, the Sox family of transcription factors may be of particular interest, as members of this family are strongly expressed in neuroectodermal progenitors as well as distinct types of glial cells of the CNS . Using mouse mutants, the three closely related group E genes SOX8, SOX9 and SOX10 (henceforth referred to as SoxE genes) have been shown to regulate glial development. SOX9, for instance, is broadly expressed in radial glia, astrocytes and oligodendrocytes of the developing CNS, and found in adult astrocytes and Bergmann glia. In the mouse spinal cord, SOX9 is required for the specification of oligodendrocytes and astrocytes from glial precursor cells . SOX10, on the other hand, is selectively expressed in the oligodendrocyte lineage during development and in the adult, and is essential for terminal oligodendrocyte differentiation and hence CNS myelination . Compared with these two SoxE proteins, SOX8 has only ancillary effects on glial specification and terminal differentiation despite the fact that it is largely co-expressed with SOX9 and SOX10 [10,11]. However, expression levels are significantly lower for SOX8 both during development and in the adult CNS .
SOX10 protein was reported to be widely expressed in human gliomas as demonstrated by immunohistochemistry . When both protein expression levels and percentage of SOX10-positive tumour cells were scored, oligodendrogliomas and low-grade diffuse astrocytomas usually presented with high expression indices, whereas anaplastic astrocytomas and glioblastomas either had low expression indices or were SOX10 negative. A second study found significant similarities between SOX9 and SOX10 protein expression in gliomas . However, neither of these studies looked for the expression of SOX9 or SOX10 in the tumours in comparison with the expression in normal brain tissue.
SOX6 expression has also been reported in human brain tumours [15,16]. SOX6 belongs, together with SOX5 and SOX13, to the group D of Sox genes [17,18]. These SoxD genes are strongly expressed in the developing CNS [19–21], and there is good evidence that SoxE and SoxD genes influence each other's expression and activity during developmental processes [20,22,23]. We thus decided to analyse and compare the expression of all SoxE and SoxD genes using real-time reverse transcription polymerase chain reaction (RT-PCR) analyses in a series of 60 gliomas of different histological types and WHO grades. Our results indicate that SOX5, SOX9 and SOX10 expression levels in the tumour tissue are equal to or lower than those in adult brain tissue. In contrast, upregulation was observed for SOX6, SOX8 and SOX13 in oligodendrogliomas and, for SOX8, additionally in low-grade diffuse astrocytomas, with particular high expression of SOX8 in oligodendroglial tumours with allelic deletions on 1p and 19q. Furthermore, all investigated Sox genes were downregulated in glioblastoma, reflecting the highly dedifferentiated state of the neoplastic cells in the most malignant type of glioma.
Materials and methods
The human glioma samples were collected at the Department of Neuropathology, Heinrich-Heine-University Düsseldorf, and analysed in an anonymized manner as approved by the local institutional review board. In total, tumours from 60 adult glioma patients were studied. All cases were histologically classified according to the WHO classification of tumours of the nervous system . The series consisted of 11 oligodendrogliomas (WHO grade II), 22 anaplastic oligodendrogliomas (WHO grade III), 7 anaplastic oligoastrocytomas (WHO grade III), 5 diffuse astrocytomas (WHO grade II), 5 anaplastic astrocytomas (WHO grade III), and 10 glioblastomas (WHO grade IV). Samples of each tumour were snap-frozen immediately after operation and stored at −80°C. A tumour cell content of at least 80% was histologically determined for each specimen.
Acquisition of nucleic acids and microsatellite analysis for allelic losses on 1p and 19q
Extraction of genomic DNA and total RNA from frozen tumour samples was performed by ultracentrifugation as described in detail elsewhere . Leucocyte DNA was purified from blood samples according to a standard protocol . The 40 oligodendroglial tumours included in this study had been investigated before for allelic losses on 1p and 19q using loss of heterozygosity (LOH) analysis at multiple microsatellite markers from 1p34-pter and 19q13 .
RNA from non-neoplastic select adult brain regions, including frontal lobe, parietal lobe, temporal lobe, occipital lobe and corpus callosum, were purchased from BioChain (Hayward, CA, USA), and RNA from adult total brain from BioChain, Stratagene (Cedar Creek, TX, USA), BD Biosciences (San Jose, CA, USA) and Clontech (Mountain View, CA, USA).
Three micrograms of total RNA from each control and tumour tissue were reverse-transcribed in a total volume of 50 μl into cDNA using random hexanucleotide primers and SuperscriptTM reverse transcriptase (Invitrogen, Karlsruhe, Germany). From each cDNA solution, 1 μl was amplified with primer pairs specific for each of the six SoxE and SoxD genes. The following primers were used: SOX5 (5′-CTC TCC ACC TTC TCC ATC TC-3′ and 5′-AAT CTC ACC AGC TGC TGG TA-3′, according to positions 2944–2963 and 3243–3224 of acc. no. NM_006940, yielding a product of 300 bp), SOX6 (5′-GGA CAG CGT TCT GTC ATC TC-3′ and 5′-CTC TTG TTC AGT CCG AGT CA-3′, according to positions 268–287 and 505–486 of acc. no. NM_033326, yielding a product of 238 bp), SOX8 (5′-TGT CTC CTT GCT GGC AGA GT-3′ and 5′-GAG CAA CGA GCA ACG TGA TG-3′, according to positions 1988–2007 and 2271–2252 of acc. no. NM_014587, yielding a product of 265 bp), SOX9 (5′-ATC AAG ACG GAG CAG CTG AG-3′ and 5′-TCA AGG TCG AGT GAG CTG TG-3′, according to positions 1561–1580 and 1902–1883 of acc. no. NM_000346, yielding a product of 342 bp), SOX10 (5′-GGA GGC TGA AGA GGC TGA CA-3′ and 5′-AGT GGT AAG GCC TCC GAT GC-3′, according to positions 1783–1802 and 2051–32 of acc. no. NM_006941, yielding a product of 269 bp), SOX13 (5′-CAG CCA GCC AAC CTA AGA CT-3′ and 5′-GTC CAG CCA GGA CAA GAA TG-3′, according to positions 2528–2547 and 2842–2823 of acc. no. NM_005686, yielding a product of 315 bp), and β-actin (5′-CCTGGGCATGGAGTCCTG-3′ and 5′-GGAGCAATGATCTTGATCTTC-3′, yielding a product of 179 bp). Polymerase chain reactions were performed on a Roche Lightcycler (Roche Diagnostics, Mannheim, Germany) according to the manufacturer's instructions using the LightCycler-FastStart DNA Master SYBR Green kit with an annealing temperature of 60°C.
Determination of relative expression levels and statistical analysis
Transcript levels for SoxE and SoxD genes were normalized to the β-actin levels in the same sample. The mean value of the normalized mRNA levels for the four non-neoplastic adult total brain tissue samples served as reference and was set to 1. Normalized mRNA levels for SoxE and SoxD genes in non-neoplastic and tumour tissues were calculated as fold increases relative to this mean value.
The mRNA levels in the examined tumour subgroups did not fit a normal (Gaussian) distribution. Therefore, the median values and ranges, rather than their mean values and coefficients of variation, were presented. The statistical significance of detected differences was evaluated using the Mann–Whitney U-test that enables the determination of equality of two variables sampled from an unknown distribution. Differences between two populations were judged either as not significantly different (P ≥ 0.05), marginally significantly different (P < 0.05), or significantly different (P < 0.01), respectively. The values obtained for select adult brain regions indicate the range in which Sox gene expression varies due to regional differences in cellular composition.
Protein extracts and Western blot analysis
Protein extracts were obtained as a by-product from RNA extraction by ultracentrifugation . Following dialysis against 0.1% Triton X100 in phosphate-buffered saline using Slide-A-Lyser MINI Dialysis Units (Pierce, Rockford, IL, USA) and determination of protein concentrations, 10–15 μg of each protein extract was run on a 12% SDS gel. Size-separated proteins were electrotransferred to a nitrocellulose membrane that was first blocked and then consecutively incubated with a polyclonal SOX8 antiserum from guinea pig  and horseradish peroxidase-coupled rabbit anti-guinea pig IgG antibodies (Zymed, San Francisco, CA, USA), both diluted 1:6000. Immunoreactive bands were visualized by autoradiography using the chemiluminescent ECL plus reagent (Amersham Pharmacia Biotech, Piscataway, NJ, USA). To control for protein amounts, membranes were stripped after SOX8 detection and reprobed by consecutive incubation with a monoclonal β-actin antibody (1:4000 dilution; Sigma, Saint Louis, MI, USA) and a horseradish peroxidase-conjugated goat anti-mouse IgG antibody (1:25 000 dilution; Pierce, Rockford, IL, USA). Immunoreactive bands were again visualized using the ECL system.
SoxE gene expression in non-neoplastic brain tissue
To get an impression of Sox gene expression in the adult brain, we not only measured the average expression level of each SoxE and SoxD gene in total brain, but also determined the range in which expression varies among different brain regions, including frontal lobe, parietal lobe, temporal lobe, occipital lobe and corpus callosum (Figures 1 and 2). As expression of SoxE and SoxD genes is restricted to different cell types in the adult brain, the observed variations represent the different cellular composition of specific brain regions. This naturally observed range of expression needs to be taken into account when interpreting the observed Sox gene expression in tumours. It is therefore highlighted by a grey shade in Figures 1 and 2.
SoxE gene expression in oligodendroglial tumours
When analysing SOX10 gene expression in oligodendrogliomas, we first realized that most tumours showed overall SOX10 expression levels below the average of non-neoplastic adult total brain (Figure 1c). Taking into account the previously reported widespread expression of SOX10 in oligodendroglioma [13,14], one may conclude that the SOX10 expression levels in tumour cells are significantly lower than in the normal oligodendrocytes of the adult brain. As we consistently observed a strong increase of SOX10 expression in developing oligodendrocytes upon terminal differentiation , the lower expression level in tumour cells likely corresponds to their less differentiated state, as supported by the fact that oligodendroglioma cells usually lack expression of major myelin proteins . Our finding of highest SOX10 expression levels in well-differentiated oligodendrogliomas would support an association between SOX10 expression and the degree of differentiation in oligodendroglial tumours (Figure 1c). SOX10 expression levels were heterogeneous in the group of WHO grade III oligodendrogliomas and anaplastic oligoastrocytomas, which may suggest differences in the differentiation state of the tumour cells or could alternatively be due to variable amounts of contaminating normal oligodendrocytes in the different tumour samples.
In agreement with previous findings , SOX9 expression in oligodendroglial tumours behaved in a very similar manner to SOX10 expression, with the mRNA levels in the tumours usually being lower than those in the average adult total brain (Figure 1b). The only exception was a single anaplastic oligodendroglioma that demonstrated a very high SOX9 mRNA level (Figure 1b).
In contrast to SOX9 and SOX10, we detected a significant upregulation of SOX8 mRNA in the vast majority of oligodendroglial tumours (Figure 1a). SOX8 expression was furthermore higher in most tumours than the upper range of region-dependent variation in the brain. Upregulation was consistently found in the WHO grade II oligodendrogliomas, while a small subset of anaplastic oligodendrogliomas lacked increased SOX8 levels (Figure 1a). Accordingly, the median of SOX8 expression was increased relative to non-neoplastic total brain to 13-fold in the WHO grade II oligodendrogliomas and to 9- and 8-fold in the anaplastic oligodendrogliomas and oligoastrocytomas. SOX8 expression was inversely related to SOX9 and SOX10 expression in oligodendrogliomas. The increased SOX8 mRNA levels in oligodendroglioma cells are in line with reports showing high SOX8 expression in immature glial cells  but fairly low levels in adult brain tissue .
SoxE gene expression in astrocytic tumours
To determine whether SOX8 upregulation is specific for oligodendrogliomas, we also analysed its expression in a series of astrocytic gliomas (Figure 1a). In low-grade diffuse astrocytomas, SOX8 expression was increased comparably to low-grade oligodendrogliomas. In contrast, anaplastic astrocytomas exhibited heterogeneous SOX8 expression, with some tumour samples showing increased and others reduced SOX8 mRNA levels (Figure 1a). Nine of 10 investigated glioblastomas demonstrated lower SOX8 levels as compared with adult total brain (Figure 1a).
SOX9 expression was also upregulated in most cases of low-grade diffuse astrocytoma, whereas anaplastic astrocytomas and glioblastomas exhibited variable expression levels (Figure 1b). SOX10 expression was downregulated in 50% of the WHO grade II and III astrocytomas, and consistently low in the glioblastomas (Figure 1c).
SoxD gene expression in oligodendroglial tumours
We have recently shown that SoxD genes are co-expressed with SoxE genes in CNS glia during normal mouse development and are capable of counteracting their ability to induce oligodendroglial specification and terminal differentiation . Therefore, we determined expression of the three SoxD genes SOX5, SOX6 and SOX13 in the same set of glioma samples that were analysed for SoxE gene expression.
The majority of WHO grade II and III oligodendrogliomas exhibited decreased SOX5 mRNA levels relative to the already low levels in adult brain tissue (Figure 2a). Anaplastic oligoastrocytomas were characterized by either modestly increased or decreased SOX5 levels compared with total brain.
In the majority of oligodendroglial tumour samples, increased expression was observed for transcripts encoding the related SOX6 and SOX13 proteins (Figure 2b,c). There was even a significant fraction in all three categories of oligodendroglial tumours, in which SOX6 or SOX13 expression levels were elevated more than 5-fold compared with the total brain control (Figure 2b,c). The significant upregulation of SOX13 in many oligodendroglial tumours is noteworthy, considering that SOX13 has been reported to be primarily expressed during development in immature neuronal precursors and selected populations of mature neurones .
SoxD gene expression in astrocytic tumours
Most astrocytic gliomas, including all anaplastic astrocytomas and glioblastomas, exhibited SOX5 expression levels below those detected in adult brain tissue (Figure 2a). The increased SOX5 levels in two WHO grade II astrocytomas were marginal and statistically not significant. SOX6 mRNA expression was downregulated in all glioblastomas and either slightly up- or downregulated in WHO grade II or III astrocytomas (Figure 2b). Only a single WHO grade II astrocytoma showed a more than 2-fold increase in SOX6 expression relative to adult brain tissue (Figure 2b). Moderate increases in expression were observed for SOX13 in approximately half of WHO grade II or III astrocytomas (Figure 2c). The other half exhibited decreased SOX13 expression, as did 9 of the 10 glioblastomas. Taking into consideration the fact that SoxD genes are expressed at only low levels in the adult CNS [21,23], the pathophysiological relevance and diagnostic value of these differences appear to be limited.
Correlation of Sox gene expression with LOH on chromosomal arms 1p and 19q in oligodendroglial tumours
As LOH on chromosomal arms 1p and 19q occurs frequently in oligodendroglial tumours and usually predicts a better response to radio- and chemotherapy as well as longer survival [2–4], we also assessed whether 1p/19q losses correlated with altered expression levels of any of the investigated SoxE and SoxD genes. Among the 40 oligodendroglial tumours included in the present series, combined deletions of 1p and 19q were detected in 8/11 WHO grade II oligodendrogliomas, 14/22 anaplastic oligodendrogliomas and 2/7 anaplastic oligoastrocytomas. Statistical analysis revealed that SOX8 mRNA expression was significantly associated with 1p/19q loss in oligodendroglial tumours (Table 1). In particular, 1p/19q-deleted tumours exhibited a significantly higher expression of SOX8 than tumours without 1p/19q loss (P = 0.0019).
Table 1. Correlations between loss of heterozygosity (LOH) on 1p and 19q and SOX8 expression in oligodendroglial tumours
LOH, loss of heterozygosity on 1p and 19q(+, present; –, absent).
Median(range): median and range of Sox gene expression levels.
P: statistical significance according to Mann–Whitney U-test.
Marginally significant(P < 0.05).
Significantly different(P < 0.01), also indicated by bold numbers.
To evaluate SOX8 protein amounts in oligodendroglial tumours, we performed Western blot analysis on 13 selected tumours and 1 non-neoplastic adult human brain sample (Figure 3). In agreement with transcript levels, SOX8 protein amounts were much higher in the oligodendroglial tumours than in the non-neoplastic adult brain tissue. Further comparison of seven randomly chosen tumours with 1p/19q deletion and six tumours without 1p/19q deletion confirmed particularly high SOX8 protein amounts in tumours with allelic losses on 1p and 19q, thus confirming the association of SOX8 and 1p/19q deletion on the protein level (Figure 3).
In this report, we performed a systematic expression analysis of all SoxE and SoxD genes in human astrocytic and oligodendroglial gliomas. Several members of the Sox gene family have already been analysed by immunohistochemistry for their presence in brain tumours. For example, it has been reported that SOX6, SOX9 and SOX10 are expressed in different types of glioma [13–16,30]. Expression of SOX10, but not SOX6, was found to correlate inversely with the tumour grade. Our data support the expression of these Sox genes in various types of glioma. However, because our study is mainly based on real-time RT-PCR data, we cannot draw any conclusions about the distribution of Sox gene expression within the tumour. Furthermore, we have not generally studied how altered gene expression levels translate into changes of protein amounts in the tumour. The determined parallel increase of SOX8 transcript and protein levels in oligodendroglial tumours and comparison with previously published data [13,15,16] indicate, however, that there is a good correlation between changes at the transcript and protein levels.
The advantage of our study, on the other hand, is that we determined expression levels for all SoxE and SoxD genes in the same set of tumour samples, and that we compared expression levels in the tumour with those in normal adult brain tissue. To interpret our data, it is thus important to consider normal expression levels for each Sox gene in adult brain tissue. Although all SoxE and SoxD genes are expressed in the adult brain, expression levels vary significantly. SOX9 and SOX10 remain strongly expressed in mature astrocytes and oligodendrocytes, respectively [8,9]. This high constitutive expression in the adult brain explains why we failed to detect upregulation of SOX9 or SOX10 transcripts in gliomas, although previous studies showed SOX9 and SOX10 protein expression in many gliomas.
SOX8 and the three SoxD genes are most strongly expressed in the developing CNS in immature cells and become downregulated to lower basal levels in the adult brain [10,12,23]. Taking these different basal expression levels in the adult brain into account, decreases appear to be mainly informative for Sox genes with high constitutive expression levels (that is, SOX9 and SOX10), whereas reductions of transcript levels for Sox genes with already low expression in normal brain, for example SOX5, should be interpreted with caution. For the latter group of Sox genes, increases in expression levels (as observed for SOX6, SOX8 and SOX13) are more meaningful.
Regional differences also have to be taken into account, as a gene with preferential expression in oligodendrocytes such as SOX10 will be more heavily expressed in white matter than in grey matter regions. Our analysis of isolated brain regions has indeed confirmed such a regional variation of expression. Nevertheless, expression levels in many tumours differed much more from average total brain values. In these cases, altered expression in the tumour cannot be masked or explained by regional differences of SOX gene expression levels.
The decreased expression levels for SOX9 and SOX10 in gliomas are as much indicative of a less differentiated state than the increased levels for SOX8. As SOX8 expression during normal development marks an immature glial state rather than a completely differentiated state , SOX8 levels may be instrumental in predicting the differentiation status of a glioma specimen, in particular in cases of WHO grade III gliomas, because tumours within this group display widely varying SOX8 levels that correspond to those observed either in the low-grade gliomas or in glioblastomas. Quite interestingly, high SOX8 transcript and protein levels in oligodendroglial tumours correlated well with LOH on 1p and 19q, suggesting that SOX8 expression may possibly be useful in predicting therapeutic response and prognosis in patients with oligodendroglial neoplasms.
SOX10 has been shown to activate myelin gene expression [9,31]. The continued presence of SOX10 in many gliomas – albeit at lower levels – is thus at odds with the low level of myelin gene expression in gliomas. During development, SoxD proteins appear to modulate SOX10 function and prevent SOX10 from activating myelin gene expression in immature oligodendrocytes . Thus, it is intriguing that oligodendrogliomas, oligoastrocytomas and WHO grade II astrocytomas often contain increased SOX6 levels. In analogy to the situation during normal development, increased SOX6 levels may keep the myelin-inducing activity of SOX10 in check and are thus also representative of a less differentiated cellular state. In its function, SOX6 may be aided by its close relative SOX13, which we also found to be expressed at elevated levels in many gliomas. SOX13 may function similarly to SOX6, as often observed for closely related SOX proteins during development . Quite interestingly, SOX13 is normally not expressed in the same cells as SOX6. Whereas SOX6 is restricted to immature glia during normal CNS development , SOX13 is primarily found in immature neuronal precursors . The joint upregulation of both SOX6 and SOX13 in gliomas, as observed in our study, is thus most easily explained by a model in which gliomas arise from neuroepithelial stem or progenitor cells. Furthermore, none of the analysed SoxE or SoxD genes exhibited an expression pattern that set oligodendroglial tumours clearly apart from astrocytic tumours, thus contradicting a model in which these tumour entities arise from differentiated oligodendrocytes and astrocytes, respectively. This finding is in line with recent reports on other oligodendroglial lineage-associated transcription factors, such as OLIG1, OLIG2, NKX2.2, HEB and E2A, which were also found to be expressed in both oligodendroglial and astrocytic tumours [32,33]. As shown in our present study, Sox gene expression, such as expression of SOX8, SOX6 and SOX13, however, may be helpful to distinguish low-grade from high-grade gliomas, in particular glioblastomas, in which Sox gene expression patterns correspond to a very immature glial differentiation state.
This work was supported by grant 2002.048.1 from the Wilhelm Sander-Stiftung and by grant 70-3088-Sa1 from the Deutsche Krebshilfe.