Subtype-specific expression and genetic alterations of the chemokinereceptor gene CXCR4 in medulloblastomas



Recent findings indicate that the chemokine receptor Cxcr4 is essential for normal development of the cerebellar cortex. As medulloblastomas (MBs), the most common malignant brain tumors of childhood, are believed to arise from neuronal cerebellar precursors, we asked whether there is a potential role for Cxcr4 in the pathogenesis of MB. RT-PCR and immunohistochemistry revealed expression of Cxcr4 in different variants of MBs. Whereas 18/20 classic MBs showed very low levels of CXCR4 mRNA, high amounts were expressed in 17/18 desmoplastic and 6/7 extensively nodular MBs. In addition, a significant correlation of high CXCR4 mRNA levels and presence of the neurotrophin receptor p75NTR or expression of ATOH1 and GLI1 suggests that CXCR4 is a reliable marker for tumors derived from the cerebellar external granular layer. Because Cxcr4 is important for migration and cell cycle control of granular precursors, we screened for mutations in the coding region by SSCP and gene sequencing. In a series of 90 MBs and 8 MB cell lines, we found one germline and one somatic mutation resulting in amino acid substitutions in the first (Ile53Leu) and second (Asp97Asn) transmembrane regions, respectively. These data suggest that Cxcr4 may be involved in the pathogenesis of MBs. © 2005 Wiley-Liss, Inc.

MBs are the most common malignant brain tumors in childhood, with about 350 new cases each year in the United States. The median age at onset is 9 years, and 5-year survival rates reach up to 79%.1 Although the majority occur as sporadic cases, hereditary conditions, including Gorlin's and Turcot's syndromes, have been associated with a significantly elevated incidence of MB.2, 3 Mutations in the patched and the APC genes, which are known to cause these syndromes, have been found in sporadic MBs. Alterations in genes of several members of the Sonic Hedgehog and the WNT/wingless signaling pathways have also been reported in MBs.4, 5, 6

According to the current WHO classification of nervous system tumors, MBs are histologically divided into 4 major subgroups, which significantly differ in histopathology, genetics, location and clinical characteristics. While classic and desmoplastic tumors account for the vast majority of cases, MBs with extensive nodularity and large cell MBs are rare.7 All MB subtypes are believed to derive from neural progenitors of the cerebellum. However, the exact cellular origin remains to be elucidated in most cases. Several lines of evidence indicate that desmoplastic tumors, which account for about one-fourth of all cases, originate from neuronal precursors in the EGL of the cerebellar cortex. Similar to the EGL, DMBs consist of highly proliferating cells. They produce a strong reticulin fiber network and contain clusters of more differentiated cells, which form reticulin-free “pale” islands. Beyond morphologic aspects, activation of the Sonic Hedgehog–patched pathway is a common and characteristic molecular feature both in granule cell development and in DMBs (for review, see Wechsler-Reya and Scott8).

The G protein–coupled 7-transmembrane chemokine receptor Cxcr4 was originally discovered in lymphocytes, where it represents a major cofactor for the fusion and entry of HIV-1.9 It has also been implicated in the development of the cerebellar cortex. Cxcr4 is strongly expressed in proliferating granule cell precursors and a target gene of the Sonic Hedgehog pathway.10, 11 SDF-1, which is the only known ligand of Cxcr4 and which is segregated by meningeal cells of the leptomeninx, significantly enhances Sonic Hedgehog–induced cell proliferation.10 This effect is reversible by blocking the Cxcr4 receptor either by a specific small-molecule inhibitor, AMD 3100, or by pertussis toxin, indicating coupling of neuronal Cxcr4 to Gαi, which has previously been demonstrated to be expressed in granule cell precursors.12, 13

The importance of both ligand and receptor for the formation of the cerebellar cortex has further been demonstrated using knockout mice lacking either SDF-1 or CXCR-4. Both mouse models have nearly the same phenotype: they suffer from severe cerebellar abnormalities with a misplaced EGL and clusters of proliferating granule cell precursors which have migrated inappropriately deep within the cerebellar anlagen. Secondary to the failure of adequate granule cell development, the overall shape of the SDF1–/– and CXCR4–/– cerebella is altered, with Purkinje cells which are ectopically located and absence of foliation.14, 15 Thus, evidence is accumulating that the SDF-1–Cxcr4 axis plays a crucial role in normal development of the cerebellar cortex and in cell cycle control of neuronal precursors within the EGL.

These findings together with observations showing Cxcr4 to be involved in growth and progression of various human cancers16, 17, 18 prompted us to investigate its role in cerebellar MBs. A large series of tumor specimens, including CMBs, DMBs and MBENs, was used to examine the expression pattern of the Cxcr4 receptor. In addition, we performed mutational screening for genetic alterations that might contribute to abnormal Cxcr4 activity in MBs.


APC, adenomatous polyposis coli; CMB, classic medulloblastoma; DMB, desmoplastic medulloblastoma; EGL, external granule cell layer; 5-FAM, 5-carboxyfluorescein; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; MAb, monoclonal antibody; MB, medulloblastoma; MBEN, medulloblastoma with extensive nodularity; SDF, stromal cell–derived factor; SSCP, single strand confirmational polymorphism.

Material and methods

Patients, tumors and cell lines

A total of 90 MB tumor biopsies and 11 MB cell lines were analyzed. MBs included 40 tumors of the classic and 43 tumors of the desmoplastic subtypes, as well as 7 MBs with extensive nodularity. All tumors were classified according to the revised WHO classification of brain tumors.7 Patients' age at diagnosis ranged from 1 month to 59 years, with a median of 11 years. The majority of patients were enrolled in the German Society of Pediatric Hematology and Oncology multicenter treatment study for pediatric malignant brain tumors. DNA was available in most cases from peripheral blood. The analyzed cell lines were DAOY,19 D283 MED,20 D425 MED,21 D341 MED,22 D556 MED,23 MHH-MED-1,24 MHH-MED-2,24 MHH-MED-3,24 MEB-MED-8A, MEB-MED-8S and 1580WÜ (T.P., unpublished). DAOY, D283 MED and D341 MED cells were purchased from the ATCC (London, UK). Control DNA was obtained from blood samples of 125 healthy volunteers without a history of cancer or any CNS disease. For mRNA expression studies, 5 adult and 5 fetal (15.5–19 weeks of gestation) human cerebellar samples were included as control tissues. For immunohistochemical analyses of Cxcr4 expression, specimens of a fetal (21 weeks of gestation) and an adult human cerebellum were obtained at autopsy. Fetal cerebella were obtained from electively aborted fetuses with the mothers' consent under a fetal tissue collection protocol approved by the hospital's ethics committee. Histologic examinations did not show signs of neurologic disorders or cancer. RNA of adult cerebella was taken from tissue adjacent to vascular malformations. Frozen sections of the tissue taken for RNA extraction did not show any alterations. Patients gave informed consent to use the material, which had to be removed during an operation.

DNA extraction, RNA extraction and cDNA preparation

DNA was isolated from peripheral blood, tumor tissue and cell lines by standard Proteinase K/SDS digestion followed by phenol/chloroform extraction. Tissue fragments selected for DNA or RNA extraction were checked by microscopy to ensure that they consisted of either tumor or normal cerebellar tissue. Only fragments with a tumor cell content of at least 80% were included. Total cellular RNA was extracted with the Trizol reagent (Invitrogen, Carlsbad, CA), according to the manufacturer's instructions. To remove any contaminating DNA, samples were digested with DNase I (Promega, Madison, WI). Subsequently, DNase was removed by an additional Trizol extraction. Reverse transcription was performed using the Superscript II Preamplification System (Invitrogen) with random hexamers as primers in a final volume of 20 μl. Contamination with genomic DNA was excluded by PCR using intron-spanning CDK4 primers (forward 5′-CAT GTA GAC CAG GAC CTA AGG-3′, reverse 5′-AGC TCG GTA CCA GAG TGT AAC-3′) in each sample.

DNA amplification and mutational analysis of the CXCR4 gene

Genomic DNA from each tumor and blood sample was amplified using 6 overlapping primer pairs. PCR was carried out in a final volume of 10 μl with 10–50 ng of genomic DNA in a buffer containing 1.0–1.5 mM MgCl2 (Invitrogen), 200 mM of each deoxynucleoside triphosphate, 5 pmol of each forward and reverse primer and 0.25 U Taq polymerase (Invitrogen). After an initial denaturation step of 94°C for 5 min, 35 cycles of PCR were performed for all primers on a Thermoblock cycler (Biometra, Göttingen, Germany) with a denaturation step of 94°C for 35 sec, an annealing step of 57°–59°C for 40 sec and an extension step of 72°C for 40 sec. Primer sequences are given in Table I. Mutational analysis was performed using the SSCP method. PCR products were loaded onto 10% or 14% polyacrylamide (acrylamide:bisacrylamide ratio 1:29, 1:59 or 1:79) gels with or without 10% glycerol and run at room temperature (60 V) or at 4°C (80 V) for 18–24 hr. Single and double strands of PCR products were visualized by silver staining as described elsewhere.25 PCR products showing aberrantly migrating bands were reamplified at least twice in independent experiments. Bands were then excised and eluted, and the DNA was reamplified. The resulting PCR products were purified using the QIAquick PCR purification kit (Qiagen, Chatsworth, CA). Sequencing was performed with the PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit and the GeneAmp PCR system 9600 (Perkin-Elmer, Foster City, CA), using 20 ng of PCR product as a template. Sequencing reaction products were separated on an Applied Biosystems (Foster City, CA) 377 sequencer.

Table I. Primer Sequences and PCR Conditions for Mutational Analysis of CXCR4
CodonDirectionPrimer sequencePCR productlength (bp)MgCl2(mM)Annealingtemperature (°C)

Semiquantitative RT-PCR

Expression of CXCR4 mRNA was assessed semiquantitatively using duplex RT-PCR for β2-microglobulin and CXCR4 in 45 cases for which total RNA was available and in 9 MB cell lines. Forward and reverse primer sequences are indicated in Table II and located in different exons spanning intronic sequences to exclude signals from residual genomic DNA. Reverse primers were 5′-labeled with fluorescent dyes (5-FAM; MWG Biotech, Ebersberg, Germany). Duplex RT-PCRs were carried out in a final volume of 10 μl containing 100 ng of cDNA, 10 pmol of each β2-microglobulin primer and 2.5 pmol of the CXCR4 primers, 10 mM TRIS-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl2 and 200 mM of each deoxynucleotide. The PCR program consisted of initial denaturation at 94°C for 5 min followed by 36 cycles of 94°C for 30 sec, 58°C for 40 sec and 72°C for 40 sec and a final extension step at 72°C for 10 min. The number of cycles was determined to be in the exponential phase by kinetic experiments using 20–38 cycles. The resulting PCR products were loaded onto 4.5% denaturing acrylamide gels and analyzed on a semiautomated DNA sequencer (ABI 377) equipped with Genescan software (Applied Biosystems). To increase statistical validity, PCRs for the whole set of cDNAs were performed in at least 3 independent experiments. Relative expression levels were calculated as ratios of the signal intensities of CXCR4 and β2-microglobulin products. Leukocytes from a healthy man were taken as a positive control and showed a mean value of 0.48 ± 0.18 for the ratio of CXCR4 and β2-microglobulin mRNA expression. The non-small cell lung cancer cell line NCI-H292, which has previously been demonstrated to be negative for Cxcr4,26 did not show any signals for CXCR4 at all. The median of all values was 0.245 and was chosen as the cut-off for strong or weak expression of CXCR4. Duplex RT-PCR was also used to investigate mRNA expression of ATOH1 and GLI1. For ATOH1, reactions were carried out with 10 pmol of each β2-microglobulin and ATOH1 primer. For GLI1, we used 20 pmol of each forward and reverse primer together with 2 pmol of GAPDH primers. All primer sequences and PCR conditions are given in Table II.

Table II. Sequences and PCR Conditions of Primers used for Semiquantitative RT-PCR
GeneDirectionPrimer sequencePCR product length (bp)MgCl2 (mM)Annealing temperature (°C)


Paraffin sections were cut at 4 μm, mounted on slides suitable for the capillary gap method (Fisher Scientific, Fair Lawn, NJ) and air-dried in an incubator overnight at 37°C. Before staining, sections were deparaffinized in xylene and rehydrated in a graded alcohol sequence. For reactions with p75NTR antibodies, sections were boiled for 2 × 5 min at 600 W in citrate buffer solution (pH 6.0) in a microwave oven. Immunostaining was carried out on a Tech Mate staining apparatus (Dako, Copenhagen, Denmark) using reagents for the indirect streptavidin-peroxidase method provided by the manufacturer. Mouse MAbs against p75NTR (Dako) and Cxcr4 (R&D Systems, Minneapolis, MN) were used at concentrations of 2.62 μg/ml and 3.33 μg/ml, respectively. Isotype controls did not show any detectable signals. Finally, sections were dehydrated in graded alcohols and mounted in Corbit-Balsam (Hecht, Kiel, Germany). A higher number of infiltrating leukocytes in tumors compared to control tissue that might contribute to the stronger expression of Cxcr4 was ruled out by immunohistochemistry using antibodies against CD45 (3.7 μg/ml, Dako).

Fluorescence-activated cell analysis

For flow cytometry, a total of 1.3 × 106 cells of each MB cell line were stained with mouse MAbs against Cxcr4 (R&D Systems, 10 μg/ml) for 45 min at 4°C. After washing and staining with Alexa 488-conjugated secondary antibodies (Molecular Probes, Eugene, OR; 0.3 μg/ml), cells were again washed 3 times and analyzed on an LSR II flow cytometer (Becton Dickinson, Mountain View, CA). Flow cytometric data were analyzed using FlowJo 3.3 software (Tree Star, San Carlos, CA).

Statistical analysis

The statistical significance of differences in RT-PCR expression pattern of fetal and adult cerebella as well as of different MB variants was analyzed using Student's t-test for dependent samples. Correlation of high or low expression of CXCR4 with expression of p75NTR, ATOH1 and GLI1 was examined using the χ2 test with Yates' correction for continuity.


Cxcr4 protein is expressed both in neuronal precursorsof cerebellar EGL as well as in cerebellar MB

We examined expression of the Cxcr4 chemokine receptor protein in the developing cerebellum as well as in surgically removed biopsy specimens of MB. Analysis of Cxcr4 expression in a 21-week gestation fetal cerebellar cortex revealed conspicuous immunoreactivity, predominantly in proliferating neuronal precursors of the EGL (Fig. 1a,b). In contrast, mature granule cells of adult cerebella did not show significant Cxcr4 expression (Fig. 1c,d). All of the 18 MB specimens expressed immunohistochemically detectable Cxcr4 protein. In the subgroup of desmoplastic tumors, protein expression tended to be stronger in highly proliferating, reticulin-rich tumor cell areas relative to more differentiated nodular islands. On average, >75% of tumor cells were positively stained in internodular areas, while only up to 50% of tumor cells within pale islands showed Cxcr4 expression (Fig. 1e,f). A higher number of leukocytes in tumors compared to control tissue that might contribute to the stronger expression of Cxcr4 was ruled out by immunohistochemistry using antibodies against CD45 (data not shown).

Figure 1.

Immunohistochemical detection of Cxcr4 in a 21-week fetal cerebellum (a,b), an adult cerebellum (c,d) and MB (e,f). While protein is strongly expressed both in neuronal precursors of the cerebellar EGL as well as in tumor cells of desmoplastic/nodular MB, it is hardly detectable in adult cerebella. (c) High-power view of boxed region in (a) shows parts of the developing cerebellar cortex. (d) Higher magnification of boxed region in (c) with adult cerebellar cortex. (f) High-power view of boxed region in (e) with typical pale islands and cell-dense, reticulin-rich internodular areas. Scale bars = 200 μm (a,c,e), 100 μm (d) and 50 μm (b,f).

Expression of CXCR4 significantly decreases duringcerebellar development and is specifically upregulatedin MB of the desmoplastic/nodular subtype

To thoroughly investigate the quantitative amounts of transcribed CXCR4, we performed RT-PCR analyses on fetal and adult cerebella as well as on different subtypes of MB. Investigation of 5 fetal and 5 adult human cerebella revealed significant differences, with high expression of CXCR4 in fetal (mean ratio of CXCR4 and β2-microglobulin mRNA expression = 2.25, SEM = 0.68) and low expression in adult (0.07 ± 0.02, p = 0.012) cerebella (Fig. 2a). CXCR4 expression in MBs highly correlated with histologic tumor subtype (Fig. 2b). Examination of 45 MB RNA samples revealed high amounts of mRNA in DMBs (1.5 ± 0.31, n = 18) and MBENs (0.87 ± 0.16, n = 7) compared to relatively low expression of CXCR4 in CMBs (0.17 ± 0.07, n = 20; p = 0.0001 and p = 0.0005, respectively). Expression levels of DMBs and MBENs did not differ to a significant extent (p = 0.24). Similar results were obtained by RT-PCR and protein expression analyses of MB cell lines, with absence or lower expression of CXCR4 in 8 lines derived from CMBs compared to strong signals in DAOY cells, which were derived from a DMB (Fig. 3).

Figure 2.

mRNA expression of Cxcr4 in normal cerebella and in MBs. (a) Expression of CXCR4 is high in fetal cerebella (n = 5) ranging 15.5–19 weeks of gestation, while adult cerebella (n = 5) show only small amounts of CXCR4 mRNA (p = 0.012). (b) Compared to CMBs (n = 20), DMBs (n = 18) and MBENs (n = 7) express significantly higher amounts of CXCR4 mRNA (p = 0.0001 and p = 0.0005, respectively). DMBs and MBENs do not show significant differences in CXCR4 expression (p = 0.2351).

Figure 3.

CXCR4 expression in MB cell lines. (a) mRNA expression levels with high amounts of CXCR4 transcript in DAOY cells and lower expression levels or absence of CXCR4 in 8 cell lines derived from CMBs. (b) Analyses of Cxcr4 protein expression in 4 different MB cell lines. Histograms depict relative fluorescence intensities of MB cells stained with anti-Cxcr4 MAbs (dotted line) compared to that of the same cells stained with secondary antibodies (solid line). Isotype controls did not show any signals above the solid histograms.

CXCR4 mRNA expression correlates with expressionof ATOH1, GLI1 and p75NTR and is a marker fortumors derived from cerebellar EGL

The p75NTR receptor as well as the ATOH1 and GLI1 transcription factors have previously been described to be expressed in the cerebellar EGL and to be markers for MBs derived from precursor cells within this layer. Indeed, their expression patterns significantly correlate with the desmoplastic subtype of MB.27, 28, 29 In view of our results concerning CXCR4 expression in normal cerebella and in MBs, we compared expression patterns of CXCR4 with other markers for EGL-derived tumors in 40 cases of MB (16 DMB, 17 CMB and 7 MBEN).

All tumors showing high amounts of CXCR4 simultaneously expressed p75NTR and vice versa (p < 0.0001). Expression of ATOH1 and GLI1 correlated in all but one case with CXCR4 mRNA levels (p < 0.001). Most intriguingly, expression was still comparable when looking at desmoplastic or nodular tumors that exceptionally expressed only low amounts of CXCR4 or at classic tumors that exceptionally showed strong expression of CXCR4 (Table III, Fig. 4).

Table III. Expression of CXCR4, p75NTR, ATOH1 and GLI1 in CMB (n = 17), DMB (n = 16) and MBEN (n = 7)
  1. +, high expression. −, low or no expression –n.d., not done.

p value (correlation with CXCR4 expression)<0.001<0.001<0.001
Figure 4.

Correlation between expression of CXCR4 and further markers for the cerebellar EGL in 2 representative cases of MB. While most desmoplastic and nodular MBs are characterized by coexpression of CXCR4, p75NTR, ATOH1 and GLI1, CMBs mostly show low expression of CXCR4 and absence of p75NTR, ATOH1 and GLI1. Peak areas indicate amount of PCR products. Scale bar = 200 μm.

Mutational analysis of CXCR4 uncovers one germline and one somatic mutation in MB patients

To screen for genetic alterations, the entire coding sequence of the CXCR4 gene was analyzed by SSCP in 98 DNA samples from 90 MB tumor biopsies and 8 MB cell lines. Silent sequence variants with a C→T substitution were found in 7 cases without any subtype association (exon 2, position 414, according to GenBank accession AF025375). These alterations were present both in the tumor DNAs as well as in the corresponding constitutional blood DNA and are the result of a single nucleotide polymorphism previously described by others.30 Analysis of one additional patient with DMB also resulted in aberrant SSCP bands from both the tumor (D978) and its matching blood (D979) DNA. In this case, gene sequencing identified an A→C substitution at position 157 in exon 2 (Fig. 5a). This substitution causes an amino acid switch from isoleucine to leucine (Ile53Leu) within the first transmembrane region of the Cxcr4 receptor (Fig. 5c). To exclude a rare polymorphism, which was not present in any of the other 97 tumor samples, we analyzed an additional 125 blood samples taken from healthy controls without any neoplastic or CNS disease. None of the analyzed DNA samples showed aberrantly migrating bands in SSCP analyses. Since polymorphisms are defined to occur in at least 1% of the population, this alteration has to be regarded as a germline mutation in the CXCR4 gene. Furthermore, we discovered a somatic mutation in the DNA of the CMB D999. Reamplification of the excised band and DNA sequencing revealed a G→A transition in codon 97 leading to a substitution of aspartic acid by asparagine in the second transmembrane region of the receptor. This alteration was not detectable in the matching leukocyte DNA of the same patient (D1000, Fig. 5b,c).

Figure 5.

Mutational analysis of CXCR4 in MB. (a) SSCP analysis showing aberrant bands in tumor (D978) as well as in matching leukocyte (D979, arrow) DNA. DNA sequencing of the excised and reamplified DNA products revealed an A→C transversion in codon 53 leading to a substitution of isoleucine by leucine in both the tumor and the corresponding peripheral blood. Other tumors or healthy controls did not show any alterations in this region. (b) SSCP analysis of a somatic point mutation showing an aberrant band of amplified DNA from the tumor sample (D999), which is not present in the parallel blood sample (D1000, arrow). DNA sequencing revealed a G→A transition in codon 97 leading to a substitution of aspartic acid to asparagine in the tumor. (c) Schematic diagram of the chemokine receptor Cxcr4 with indications of the positions of the mutated residues in the first (Ile53Leu) and second (Asp97Asn) transmembrane regions.


Although MBs represent the most frequent malignant brain tumors in childhood, the molecular basis for their pathogenesis is only partially understood. In this study, we provide evidence that a small subset of MBs carries mutations in the gene encoding the Cxcr4 chemokine receptor. Moreover, strong expression of CXCR4 mRNA was demonstrated in MBs that likely derive from the cerebellar EGL. These data suggest that CXCR4 may be responsible for the development of specific MB subtypes.

Several genes, including target genes of the Sonic Hedgehog pathway such as GLI1, N-myc, Igf2 and bcl-2, have previously been reported to be overexpressed in MBs and to be significantly correlated with the desmoplastic variant of MB.29, 31 Although Cxcr4 has been clearly shown to play an essential role during the development of the cerebellar cortex15 and it was only recently identified as a target gene of the Hedgehog pathway,11 a study of its expression in MB subtypes has not been reported to date. Nevertheless, a possible functional relevance of the SDF-1–Cxcr4 axis in MBs was implied by Rubin et al.,12 who used the DAOY cell line for in vitro experiments and xenograft tumor models. As this line is known to be derived from a DMB,19 their results are in line with our observations showing that desmoplastic tumors express high amounts of CXCR4 mRNA.

MBENs, which have also been designated as cerebellar neuroblastoma, were first described in 1999 with a preferential occurrence in very young children, a peculiar grapelike appearance on neuroimaging and an apparently favorable outcome.32 To our knowledge, mRNA expression patterns have not been evaluated before in MBENs. We show here that expression of CXCR4, ATOH1 and GLI1 in this MB subtype is similar to that in DMBs. Together with the finding of strong p75NTR expression in MBEN (data not shown), this clearly argues for the hypothesis that both nodular and desmoplastic tumors have one common cellular origin, deriving from the cerebellar EGL. This is further strengthened by genetic similarities, such as Patched mutations, which can occur in both variants (T.P. and M.L.G., unpublished observations), and patients suffering from Gorlin's syndrome, who were observed to develop either desmoplastic or extensively nodular tumors.33 Thus, both MB subtypes may be related even though age at diagnosis and clinical outcome are different.

A challenge of MB research is the question of the cellular origin of MB tumor cells. p75NTR, ATOH1 and GLI1 are well-characterized markers for MBs that have derived from the cerebellar EGL; they are present in EGL precursors as well as in DMBs and MBENs. Interestingly, expression of these markers is also found in some tumors that have classic histology without nodules or a desmoplastic component. This is in line with observations of elevated GLI1 expression in up to 50% of MBs, including the classic variant.34 We therefore hypothesize that there are some MBs which do not develop reticulin fibers, although they have derived from the cerebellar EGL. Together with the present data describing a perfect correlation of CXCR4 with p75NTR, ATOH1 and GLI1, we further suggest that expression of CXCR4 is a better marker of tumors derived from the cerebellar EGL than classical histology and silver staining.

Screening for genetic alterations of the CXCR4 receptor gene in cancer has not been reported to date, though several lines of evidence describe a pathologically activated SDF-1–Cxcr4 axis to be significantly associated with increased proliferation or enhanced metastatic potential in a variety of human neoplasms, such as breast, prostate and small cell lung cancers.16, 17, 35 In the present study, we describe 2 cases of MBs carrying mutations in the CXCR4 gene; and it is tempting to speculate whether development and progression of tumor entities other than MB might also be associated with CXCR4 mutations. Most functional investigations of the receptor structure refer to its role in leukocytes or HIV pathogenesis. The mutations in our MB samples were found in regions distant from the cytoplasmic tail, which is impaired in patients with autosomal recessive WHIM syndrome,36 or from the second extracellular loop of the Cxcr4 receptor, which is important for the coreceptor activity of Cxcr4.37 The A157C mutation was located in the first transmembrane section and resulted in an amino acid change from isoleucine to leucine (Ile53Leu) within a domain that is highly conserved in evolution. The C414T mutation was found to affect codon 97 with a substitution of negatively charged Asp by neutral Asn. This codon is located in the second transmembrane region, a part of the receptor which is relatively close to the cell surface and possibly important for binding of its ligand, SDF-1.38 Although functional studies of these residues have not been carried out in MB cells, mutations of Asp to Asn have been described to strongly impair inhibition of Cxcr4 by the AMD 3100 bicyclam.39 Furthermore, alterations of Asp97 have been shown to result in protein stabilization and to strongly increase cell surface expression of Cxcr4.40 We therefore propose that mutations within the first and second transmembrane regions might contribute to pathologic receptor activity or to resistance to inhibitors such as AMD 3100. Nevertheless, further studies are needed to characterize the role of Cxcr4 in MBs and the cell type-specific functional consequence of its mutations.


We thank Dr. K. Schilling for helpful discussions as well as Ms. I. Dani and Mr. U. Klatt for excellent technical assistance. We are also indebted to Dr. G. Riggins and Dr. D. Bigner for providing D425 MED and D556 MED cells. NCI-H292 cells were a kind gift of Dr. J. Burger. Furthermore, we acknowledge the assistance of Dr. E. Endl (Flow Cytometry Core Facility, Institute of Molecular Medicine and Experimental Immunology, University of Bonn, Bonn, Germany).