Induction of IgG antibodies against the GD2 carbohydrate tumor antigen by vaccination with peptide mimotopes



The disialoganglioside GD2, a carbohydrate antigen, is expressed on all tumors of neuroectodermal origin, including melanoma, neuroblastoma, sarcoma and small cell lung cancer. Due to its specific expression on tumor surfaces, GD2 is an attractive target for immunotherapies. The mouse/human chimeric anti-GD2 mAb ch14.18 is already applied in melanoma and neuroblastoma trials as a passive immunotherapy. To establish an active immunotherapy alternative, we aimed to replace the poorly immunogenic ganglioside with immunogenic peptides. Previously, we used the ch14.18 antibody to select GD2 peptide mimics from a phage display library. In the present study, two mimics of the ch14.18 epitope were coupled to keyhole limpet hemocyanin and used for immunizing BALB/c mice. Induction of a specific humoral immune response towards the original antigen GD2, both purified and expressed on neuroblastoma and melanoma cells, could be demonstrated in ELISA, Western blot, and immunofluorohistochemistry. As the elicited antibodies were of the IgG isotype, the mimotope conjugates were capable of recruiting T cell help and inducing memory phenomena. In conclusion, we show that an epitope of the carbohydrate antigen GD2 can successfully be translated into immunogenic peptide mimotopes. Our immunization experiments indicate that GD2 mimotopes are suitable for active immunotherapy of GD2-expressing tumors.


synthetic C-DGGWLSKGSW-C


synthetic C-GRLKMVPDLE-C


human anti-mouse antibodies


keyhole limpet hemocyanin


tumor-associated carbohydrate antigen


Tumors expressing high levels of certain tumor-associated carbohydrate antigens (TACA) exhibit greater metastasis and progression rates 1. Well-documented examples of such TACA are GM2, GD2, and GD3 gangliosides in neuroectodermal tumors 2, 3. Their surface localization makes these gangliosides effective targets for active and passive antibody therapies. The disialoganglioside GD2 is considered especially interesting, as it is expressed on all tumors of neuroectodermal origin, including melanoma, neuroblastoma, sarcoma and small cell lung cancer 4, 5, but is absent in normal tissues and only minimally expressed in brain 3 and peripheral nerves 5, 6. Due to its localization in focal adhesion plaques at the interface of tumor cells and their substratum, it is thought to play a role in cell attachment and metastasis 7.

Several mAb against GD2 have been raised, with 14.18 8 and 3F8 9 being the most prominent examples. mAb 3F8 has been given to neuroblastoma patients in its murine form 10, and from these studies it was seen that the antigen GD2 is rarely lost following mAb treatment 11. mAb 14.18 has been further developed into a human/mouse chimeric derivative, ch14.18. Both 14.18 and ch14.18 trigger tumor cell lysis via antibody-dependent cellular cytotoxicity and complement-dependent cytotoxicity 12, 13. In preclinical models, the 14.18 antibodies have been shown to prevent the outgrowth of experimental melanoma 14 and neuroblastoma 15 tumors. Consequently, ch14.18 was applied in several clinical trials against melanoma 16, neuroblastoma 17, 18 and also osteosarcoma 18 with an acceptable safety profile. The effects of these passive antibody therapies were thought to be enhanced by a combination with systemic cytokines, GM-CSF 19, 20 and IL-2 21. In the next step, ch14.18 was expressed as a fusion protein with IL-2, to target the cytokine directly to the tumor site 22. This compound is already being tested in clinical trials 23.

As an alternative to these passive antibody applications, active immunizations against GD2 have been investigated. However, the antigen itself cannot be used as an effective immunogen due to its glycolipid nature. Carbohydrates are mostly T cell-independent antigens, and as no T cell help is involved, relatively low antibody titers and no memory responses are induced. Therefore, GD2 has to be antigenically enhanced for immunization purposes. One possibility is coupling of the ganglioside to an immunogenic carrier 24; another is the translation of the carbohydrate antigen into a mimicking peptide 25 or protein 2628 structure. Our group has generated circular decapeptide mimics of the 14.18 epitope on the GD2 antigen using the phage display technique described in a previous study 25. They were proven to be true epitope mimics (so called mimotopes) by mimicry tests in the ELISA format and by 3-D computer modeling. The aim of the present study was to evaluate the mimotopes for their capability of inducing a GD2-specific humoral immune response.


The vaccine constructs

Out of the 13 mimotopes generated for the ch14.18 epitope, the two cyclic peptides showing the best mimicry results 25, C-GRLKMVPDLE-C and C-DGGWLSKGSW-C, were selected for immunization studies. For this, they were produced synthetically. Before proceeding to immunizations, the synthetic peptides were again checked for specific recognition by the selecting antibody. Indeed, synthetic C-GRLKMVPDLE-C and C-DGGWLSKGSW-C (subsequently referred to as “GRL” and “DGG”), circularized by disulfide bridging between the flanking cysteines, were recognized by ch14.18, but not by the isotype control antibody (Fig. 1, left panel). To ensure immunogenicity and availability of sufficient T helper epitopes, the mimotopes were coupled to the immunogenic carrier, keyhole limpet hemocyanin (KLH). Specific mimicry of the ch14.18 epitope through mimotope conformation was also preserved in the conjugated form. KLH itself was only nonspecifically recognized by ch14.18 (Fig. 1, right panel).

Figure 1.

Synthetic peptide mimics of the ch14.18 epitope on disialogangliside GD2 are specifically recognized by ch14.18. Left panel: Cyclic peptides GRL and DGG, dotted in triplicates, are recognized by ch14.18, but not by the isotype control antibody cetuximab, or detecting anti-human IgG antibody (buffer control). Right panel: After conjugation to the immunogenic carrier molecule KLH, cyclic GRL and DGG mimotope conformation is preserved. ch14.18 recognized the mimotope conjugates (GRL-KLH and DGG-KLH), but only nonspecifically interacts with the carrier KLH. Controls are again clear.

Selected GD2 mimotopes elicit cross-reactive antibodies

Three groups of mice were immunized four times with either GRL-KLH, DGG-KLH, or KLH alone. Subsequently, blood was drawn and IgG serum levels were determined. To enable comparison between the groups, in a first step, antibodies directed against the carrier molecule, KLH, were measured in the ELISA format. All three groups showed comparable anti-KLH values (OD 450–570 nm 0.388 ± 0.004, 0.396 ± 0.005, and 0.418 ± 0.012 for mice immunized with GRL-KLH, DGG-KLH, and KLH, respectively), indicating successful immunization in all three groups. We then proceeded to determine the anti-mimotope titers. Uncoupled mimotope peptides, a control peptide, and the carrier protein were dotted onto a nitrocellulose membrane, and incubated with serial dilutions of sera from mimotope-KLH-immunized mice. As seen in the ELISA assay, anti-KLH titers were comparable for both groups (Fig. 2, KLH panels). Mimotope-induced antibodies were found to be cross-reactive for both peptides, i.e., GRL-KLH-immunized mice recognized GRL as well as DGG, and DGG-KLH-immunized mice recognized DGG as well as GRL (Fig. 2, GRL and DGG panels). As both peptides, although different in sequence, were characterized to be 3-D mimics of the same epitope, this finding is not surprising, but supports their structural equivalence. Interestingly, both GRL-KLH- and DGG-KLH-immunized mice reacted more strongly with the GRL peptide. Neither group recognized a cyclic decamer control peptide (Fig. 2, bottom panels), so the induced antibodies are demonstrated to be specific for the GD2 mimotopes.

Figure 2.

Sera of mice immunized with mimotope conjugates are cross-reactive with both peptides mimicking the ch14.18 epitope on GD2. Mice were immunized with GRL-KLH or DGG-KLH. Total Ig titer determinations by DotBlot against the peptides GRL and DGG, against the carrier KLH, and against a control peptide are shown. Serum dilutions are given in the figure.

Mimotope-induced antibodies recognize the original antigen GD2

To assess whether the induced antibodies not only react with the actual immunogen, but also with the original antigen, GD2, a ganglioside ELISA was performed. Both mimotope-KLH conjugates induced antibodies that reacted specifically with GD2, but not with the control ganglioside GM1 (Fig. 3). This difference in recognition was found to be statistically significant (p<0.001 for GRL-KLH-immunized mice, and p<0.05 for DGG-KLH-immunized mice). Mice immunized with the carrier KLH alone did not recognize either ganglioside, and the difference in GD2 recognition between mimotope and carrier groups again was found to be significant (p<0.001 for GRL-KLH, and p<0.01 for DGG-KLH). We thus demonstrated that immunization with the selected GD2 mimotopes indeed induces a specific anti-GD2 immune response. Ig subclasses and exact amounts of anti-GD2 antibodies (shown in Table 1) elicited by mimotope vaccination were also determined by ganglioside ELISA. In serum dilution series, IgM could be detected up to dilutions of 1:3000 in both mimotope groups, IgG1 up to 1:300 in the GRL-KLH and up to 1:1000 in the GDD-KLH group, IgG2a up to 1:1000 in both groups, and IgG2b up to 1:300 and 1:100, respectively. No IgG3 or IgA could be detected in the sera.

Figure 3.

Mimotope-induced antibodies recognize the original antigen, disialoganglioside GD2. Mice immunized with GRL-KLH or DGG-KLH developed IgG antibodies specifically reacting with GD2, but not with the control ganglioside GM1, in ELISA. KLH-immunized mice show no reactivity with either ganglioside. Box plots show median, 25th and 75th percentiles of IgG levels of respective groups. Both differences of mimotope groups in GD2/GM1 recognition, as well as differences between mimotope and control groups in GD2 recognition, are statistically significant (p<0.05) by two-tailed Student's t-test.

Table 1. Anti-GD2 antibody quantities (µg/mL) in mouse sera after mimotope immunization
Antibody subclassMimotope used for immunization

Mimotope-induced antibodies recognize GD2 in melanoma and neuroblastoma cell extracts and on the cell surface

To further confirm the relevance of the mimotope-induced antibodies, we sought to show that they also recognize GD2 on tumor cells. Cell extracts of GD2-expressing M21 melanoma cells and SK-NA-S neuroblastoma cells were prepared and separated electrophoretically. In both extracts, ch14.18, which was applied as a positive control, and sera from mimotope-immunized mice recognized GD2 at a band at 55 kDa in Western blots. Binding intensity was found to be dose dependent. Serum from mice immunized with the carrier KLH alone did not recognize the antigen (Fig. 4, results for the SK-NA-S cell extract). After deglycosylation of the blots, the 55-kDa band was no longer detectable with ch14.18 or the immune sera, indicating that the antibodies indeed recognize the carbohydrate moiety of GD2. These data show that the mimotope-induced antibodies recognize GD2 not only in the purified preparation, which had also been used in characterizing the mimotopes 25, but also in its native form.

Figure 4.

Immunodetection of GD2 by anti-mimotope sera in a Western blot of a SK-N-AS cell lysate. Blotted GD2 is detected by ch14.18 as a band at 55 kDa. Sera from mice immunized with the GRL-KLH and DGG-KLH conjugates also recognize GD2 (serum dilutions given in figure, total Ig detected), whereas sera from mice immunized with the carrier KLH alone show no reactivity with the antigen.

To demonstrate whether anti-mimotope antibodies also react with GD2 on cell surfaces, we performed immunofluorescence staining of M21 and SK-NA-S cells. mAb ch14.18 was used as a positive control (Fig. 5A, results of the M21 staining). Sera of mice immunized with GRL-KLH (Fig. 5B) or DGG-KLH (Fig. 5C) showed marked membranous staining of GD2-expressing cells. As seen in the GD2 ELISA, anti-GRL-KLH antibodies reacted slightly more strongly with the target cells. Only background staining was observed when serum antibodies from mice immunized with KLH alone were tested (Fig. 5D). SW480 control cells did not stain with any antibody (data not shown).

Figure 5.

Immunofluorescence staining of GD2 on M21 melanoma cells. Anti-GD2 IgG antibodies were detected by FITC-conjugated secondary antibodies. Nuclei were stained with Hoechst dye. Cells were viewed with a Zeiss Axioplan 2 fluorescence microscope. (A) Anti-GD2 mAb ch14.18 (positive control). (B) Sera from mice immunized with GRL-KLH. (C) Sera from mice immunized with DGG-KLH. (D) Sera from mice immunized with KLH alone do not specifically recognize GD2.


In the present study, we show that an epitope of the carbohydrate antigen GD2 can be successfully translated into immunogenic peptide epitope mimics. Disialoganglioside GD2 belongs to the class of TACA, and is expressed on malignancies of neuroectodermal origin. For these tumors, such as melanoma and neuroblastoma, therapeutic options in advanced disease are limited. Therefore, novel treatment modalities are being sought. GD2 has so far been exploited as a passive immunotherapy target with the chimeric mAb ch14.18 in neuroblastoma and melanoma trials. However, passive antibody applications have several drawbacks. First, these antibodies need to be engineered to be human/mouse chimeric or preferably fully humanized, to prevent formation of human anti-mouse antibodies (HAMA) in the patient. Also with ch14.18, which is a human/mouse chimeric antibody, the development of human anti-chimeric antibodies has been seen 20. Second, ‘artificial’ antibodies have to be repeatedly administered to achieve continuous serum levels that are needed for anti-tumor efficacy. This dependency on weekly infusions is extremely demanding for cancer patients. Third, the production costs of the amounts of chimeric or humanized antibodies needed for passive immunotherapy are enormous. For all these reasons, an active immunotherapy that would induce anti-GD2 antibodies would be an attractive solution, both circumventing multiple infusions, as well as the danger of inducing an immune response against the non-human parts of the artificial antibodies. The continuous availability of ‘natural’, polyclonal antibodies induced by active immunization is expected to show all the beneficial properties of passive immunotherapy, with the additional advantage that the induced antibodies are not of a single given antibody subclass. By being both polyclonal and of different isotypes, these antibodies are expected to be capable of triggering the whole array of antibody-mediated immune functions against the tumor.

Indeed, active immunizations against GD2 have been pursued for a long time. In initial studies, gangliosides were used directly as immunogens 29, 30, a strategy that encountered problems as carbohydrates are intrinsically T cell-independent antigens, which only elicit weak immune responses and mostly antibodies of the IgM subclass. To overcome these limitations, the weakly immunogenic gangliosides were coupled to the immunogenic carrier molecule KLH 24. Currently, a GD2-KLH conjugate vaccine is being tested in patients with malignant melanoma 31.

Yet other approaches seek to replace the carbohydrate moiety with its immunological drawbacks altogether. The first possibility to mimic carbohydrates with a protein structure was the development of anti-idiotypic antibodies (anti-Id). Indeed, several anti-Id mimicking GD2 have been generated 2628, and one of them has already been used as a vaccine in advanced melanoma patients 32. However, anti-Id are still mAb, and, when applied as antigens, can give rise to HAMA. So either they need to be humanized, or produced as recombinant Fab without the Fc portion. This limits their practicability for vaccine formulation.

Another possibility to generate mimics of any structure recognized by an antibody is the phage display technique 33. It is a technique to define peptides mimicking natural epitopes, including carbohydrate structures 3436. Phage display peptide libraries consist of filamentous phage particles displaying random peptides of defined length on their surface. By biopanning 37 such phage display libraries with an antibody of interest, epitope mimics, so-called mimotopes, can be selected. These peptides mimic the original antigen by virtue of their 3-D structure, in which the amino acids of the peptide adopt the same shape as the epitope recognized by the selecting antibody 38, 39. Thus, the poorly immunogenic carbohydrate antigens can be converted into peptide epitope mimics, which have a much greater potential of both antigenicity and memory induction. Another strength of vaccinations with these epitope mimics is that by virtue of their not being identical to the original antigen, immunological tolerance, which exists for self antigens, can be circumvented. Additionally, in the vaccine formulation, a highly immunogenic carrier molecule is included, providing epitopes for eliciting T cell help. Moreover, the vaccine is administered with an immunological adjuvant. Taking these three strategies together, immunological tolerance to GD2 can be overcome.

The mimotope strategy has already been successfully pursued for a number of carbohydrate antigens, including several TACA 35, 4042, with the resulting mimotopes being used for immunizations either coupled to KLH 43, or encoded as minigenes in a DNA vaccine approach 44. In a previous study, our group defined such mimotopes of the epitope recognized by mAb ch14.18 on GD2 25. These were, to the best of our knowledge, the first peptide mimics generated for this ganglioside antigen. Early last year, another group described the definition of a GD2 mimotope 45, which they encoded in a minigene for a DNA vaccination approach. In the present study, we aimed to assess whether ‘our’ cyclic mimotopes were immunogenic and whether the induced antibodies recognize the original antigen GD2. To ensure correct cyclization and thus peptide conformation, we chose to have them produced synthetically, and coupled to KLH. We were able to show that two selected GD2 epitope mimics elicited a humoral immune response. The induced antibodies were cross-reactive for the two peptides, a finding that strengthens the mimotope hypothesis. Obviously, both peptides, although different in sequence, share the same 3-D characteristics, which mimic the ch14.18 epitope. Importantly, the anti-mimotope antibodies confirmed this mimicry by specifically recognizing the original antigen, GD2 in ELISA, Western Blot, and on the cell surface. We thus demonstrated that the GD2 ganglioside antigen can be translated into peptide mimotopes for vaccination purposes.

Peptide vaccines that can be synthesized custom-made are cost-effective, simple to produce, and can easily be quality-controlled during the manufacturing process. They are chemically stable and contain no oncogenic, toxic, or infectious material. As active immunogens, synthetic mimotopes can continuously induce available tumor-targeting antibodies. These characteristics answer the points raised as disadvantages of passive antibody therapies, i.e., high costs and multiple necessary infusions. Additionally, the induction of undesired antibodies (e.g., HAMA) is ruled out, as not only a pure antigen, but a specific epitope is applied. Moreover, the resulting ‘natural’ humoral and cellular immune responses against GD2, compared to passive anti-GD2 immunotherapy, may even mediate enhanced anti-tumor effects.

In conclusion, we have shown that an epitope of a carbohydrate antigen can successfully be translated into immunogenic peptide mimotopes. Furthermore, immunizations with these epitope mimics induced antibodies again recognizing the original carbohydrate, in our case the disialoganglioside GD2. As the elicited antibodies were of the IgG isotype, we could show that with repeated vaccinations the mimotope conjugates are capable of recruiting T cell help and inducing memory phenomena. We thus provide evidence that GD2 mimotopes are suitable candidates for active immunotherapy of GD2-expressing tumors, such as melanoma and neuroblastoma.

Materials and methods

Cell lines and cell lysates

Two GD2-positive cell lines were used in this study, the human melanoma cell line M21 and the human neuroblastoma cell line SK-NA-S. The human colon adenocarcinoma cell line SW480, which is GD2 negative, was employed as a negative control. All cells were grown in RPMI medium (Gibco BRL, Inchinnan, UK) supplemented with 10% fetal calf serum, 1% glutamine, and 1% penicillin/streptomycin.

Total cell lysates were prepared as described previously 38, with a lysis buffer containing 20 mM Tris, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100 and a protease inhibitor cocktail (Complete, Roche, Basel, Switzerland), at pH 7.5. The protein concentration was determined photometrically, using bicinchoninic acid (BCA Protein Assay Kit, Pierce, Rockford, IL). Extracts were aliquoted and stored at –80°C.

Monoclonal antibodies

ch14.18, a mouse/human IgG1 mAb, directed against disialogangliside GD2, was kindly provided by Dr. S. D. Gillies (EMD-Lexigen, Lexington, MA). Cetuximab, also a mouse/human chimeric IgG1 mAb, directed against EGFR, was used as an isotype control.

Synthesis of vaccine constructs and DotBlot assay

The 1,12-cyclic peptides C-GRLKMVPDLE-C (GRL) and C-DGGWLSKGSW-C (DCC) were manufactured synthetically (piChem, Graz, Austria). Conformational accuracy was verified with ch14.18 in a DotBlot assay. In brief, the peptides were solubilized in PBS/20% dimethylformamide, and dotted onto a nitrocellulose membrane at 1 mg/mL. Blot strips were incubated with ch14.18 or cetuximab, respectively, and bound antibody was detected by peroxidase-conjugated anti-human IgG (Jackson ImmunoResearch, West grove, PA) using the ECL+ chemiluminescence detection protocol (Amersham Pharmacia Biotech, Little Chalfont, UK). Subsequently, the peptides were coupled via a linker (GPGPG) and S-acetyl-thio-acetate on their C terminus to a succinimidyl-4-(N-maleinimidomethyl)cyclohexan-1-carboxylate-activated immunogenic carrier, KLH. The conjugate was again checked for ch14.18 binding capability in a DotBlot assay (as above).

Immunization of BALB/c mice

Three groups (n=8) of BALB/c mice (Charles River Laboratories, Sulzfeld, Germany) were immunized i.p. with 10 µg of the mimotope conjugates, GRL-KLH and DGG-KLH, or the carrier protein KLH alone, respectively, on days 1, 15, 36 and 57. Aluminum hydroxide was used as an adjuvant in all groups. Blood was taken from the tail vein on days 0 (preimmune serum), 22, 43 and 64. Mice were treated according to European Union Rules of Animal Care, with permission no. 66.009/35-BrGT/P2004 from the Austrian Federal Ministry of Education, Science and Culture.

Titer determination

All titers were determined using the third immune sera (collected on day 64). Anti-carrier titers were assessed by ELISA. For anti-carrier titers, ELISA plates (Nunc, Roskilde, Denmark) were coated with KLH, 1 µg/mL in bicarbonate buffer, pH 9.6, by overnight incubation at 4°C. Plates were then washed with PBS/0.05% Tween 20, and nonspecific binding was blocked by incubation with PBS/1% dry milk powder. Sera were added at a dilution of 1:100 in PBS/0.01% dry milk. Bound antibodies were detected with a peroxidase-conjugated rat-anti-mouse IgG antibody (Jackson Immuno Research Laboratories, Inc., West Grove, PA). The reaction was developed with TMB substrate (BD Biosciences, San Diego, CA). OD was measured in an ELISA reader (Dynatech, Denkendorf, Germany) at 450–630 nm.

Anti-peptide titers were determined by incubation of serial dilutions of pooled sera (1:5000; 1:10 000; 1:50 000; 1:100 000; 1:1 000 000) with dotted mimotope peptides, a control peptide, and KLH. Bound antibodies were detected with a peroxidase-conjugated sheep-anti-mouse Ig antibody (Amersham) as in the DotBlot assay described above.

Ganglioside ELISA

Purified gangliosides GD2 and GM1 (Sigma, St. Louis, MO) were coated at a concentration of 1 µg/mL to ELISA plates (Nunc) by dilution in ethanol and subsequent evaporation of the diluent. Plates were washed and blocked as described above. Third immune sera were added in a dilution series of 1:100–1:3000 in PBS/0.01% dry milk, and bound antibodies were detected with a peroxidase-conjugated rat-anti-mouse IgG antibody (Jackson). For subclass determinations, mouse Ig standard dilution series were coated onto the same ELISA plates. Third immune sera were added to the ganglioside-coated wells as described above, while standards were incubated with PBS/0.01% dry milk. Rat anti-mouse Ig subclass antibodies were then added to all wells at a dilution of 1:500 in PBS/0.01% dry milk, and bound antibodies detected with a peroxidase-conjugated mouse anti-rat IgG antibody (Jackson). The peroxidase/substratum reactions were developed and measured as described above.

Cell lysate Western blots

M21 and SK-N-AS cell lysates (prepared as described above) were separated by SDS-PAGE, and blots incubated with ch14.18 or serum pool dilutions as detecting antibodies. Bound ch14.18 was detected by a peroxidase-conjugated goat anti-human IgG antibody (Jackson), and bound mouse antibodies by a peroxidase-conjugated sheep anti-mouse Ig antibody (Amersham), using the ECL+ chemiluminescence detection protocol (Amersham) and Biomax-MS films (Kodak).


M21 and SK-NA-S cells were plated at 5 × 104 cells/mL on eight-well Lab-Tek tissue culture chamber slides (Miles Laboratories Inc., Naperville, IL). SW480 cells were used as negative controls. Cells were grown overnight until half-confluent. Chamber slides were then cooled to 4°C, and washed with ice-cold PBS. Cells were fixed with 4% paraformaldehyde in PBS for 30 min, and chamber slides were incubated with 50 mM NH4Cl in PBS to quench fixation, and blocked with 1% BSA in PBS. Subsequently, cells were incubated with ch14.18 (positive control), or with pooled anti-GRL-KLH and anti-DGG-KLH third immune sera, and bound antibodies were detected by FITC-conjugated goat anti-mouse IgG (Caltag Laboratories, Burlingame, CA). Pooled third immune sera from the KLH-immunized mice were used as control. Nuclei were stained with 0.1 µg/mL Hoechst dye (Sigma) in PBS for 10 min. Cells were mounted in Mowiol mounting medium and viewed with a Zeiss Axioplan 2 (Carl Zeiss, Jena, Germany).


The work was supported by BioLife Science GmbH, Vienna, Austria; by project grant no. 10965 of the Austrian National Bank Science Fund; and by the Center of Excellence in Clinical and Experimental Oncology (CLEXO), Austrian Federal Ministry of Education, Science and Culture (GZ 200.062/2-VI/1/2002). A.B. Riemer and K.H. Brämswig are recipients of Hans & Blanca Moser Fund scholarships. The work was also supported by DFG (Lo 635–2) and Fördergesellschaft Kinderkrebs-Neuroblastomforschung to H.N. Lode. We thank Harald Kurz and Silke Gruber for excellent technical assistance.


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5