Main

Central primitive neuroectodermal tumors (cPNET) present as a heterogeneous group of neoplasias throughout the neuraxis(1). This group of tumors comprises a variety of different histologic subtypes (classic and desmoplastic medulloblastoma, melanotic medulloblastoma, medullomyoblastoma, central neuroblastoma, and pineoblastoma), the most common of which is the medulloblastoma(2).

Despite being the most common malignant brain tumor in children, comparatively little is known about the origin and molecular pathology of medulloblastoma and other PNET of the CNS. The term medulloblastoma was coined by Bailey and Cushing(3) who suggested that this neoplasm had one unique cell of origin, the medulloblast, related to an undifferentiated cell, which had first been described by Schaper(4). Raaf and Kernohan speculated that germinal cells in the external granular layer of the cerebellum gave rise to medulloblastomas(5). Hart and Earle(6) described an entity of supratentorial brain tumors that could not be distinguished histologically from classic medulloblastomas, which they called PNET. This term has been controversial ever since its first description because it encompasses tumors that can occur either infratentorial (cerebellum, brain stem, spinal cord) or supratentorial (cerebral hemispheres). PNET can occur in either the CNS (cPNET) or in peripheral soft tissues and bone (peripheral PNET). Recent evidence suggests that supratentorial PNET may differ from medulloblastoma on a molecular level. Burnet et al.(7) demonstrated loss of heterozygosity on chromosome 17p in approximately 50% of medulloblastomas but in none of the supratentorial PNET under investigation.

Peripheral PNET are classified with the ESFT because of characteristic identical genetic alterations, including expression of the pseudoautosomal MIC-2 gene and the translocation t(11;22)(8). Peripheral PNET are found at the more differentiated end of the ESFT spectrum, whereas classic Ewing's tumors are characteristically undifferentiated. Atypical Ewing's sarcomas, including extraosseous Ewing's sarcomas, represent an intermediate of these two(9).

Neuropeptides, such as somatostatin, neuropeptide Y, and VIP, have been identified very early in neural development and in developmental tumors of the CNS(10–14). These peptides appear to regulate both proliferation and differentiation of neuronal precursors. Somatostatin appears to play a crucial role in the development of the human cerebellum. Bateman et al.(15) and Gould et al.(12) demonstrated the presence of somatostatin in cPNET by immunohistochemistry, suggesting a neuroendocrine origin for these neoplasias. Leroux et al.(16) then demonstrated that somatostatin determined the morphogenesis of the rat cerebellum by an inhibition of neuroblast migration.

Somatostatin exerts its biologic effect through five different G protein-coupled receptors (sst1-5)(17). Highest concentrations of sst in the normal human brain are found in the cortex, limbic system, and discrete nuclei of the diencephalon and mesencephalon. According to data presented by Laquerriere et al.(18), the concentration of sst in the cerebellum reaches a maximum at birth and declines after final organization into the mature three-layer pattern (approximately 8 mo). The EGL contains high densities of somatostatin receptors. Further characterization of somatostatin receptor subtypes in the developing human brain has not been performed to date.

Antiproliferative actions of somatostatin have been demonstrated in a variety of adult malignancies(19–24). Somatostatin has been used as an antiproliferative agent owing to its physiologic counterregulatory functions to GH, IGF-1, and epidermal growth factor. This growth inhibition has been demonstrated in a variety of cell lines and implemented into clinical practice with somatostatin analog therapy in several IGF-1-sensitive malignancies, including breast and prostate cancers, osteogenic sarcoma, and neuroblastoma(25–27). The antiproliferative effects of SS-14 are mediated by several mechanisms, including induction of apoptosis, inhibition of growth factors, inhibition of gene expression, and inhibition of tumor angiogenesis(28,29). We have previously demonstrated the high expression of SS-14 and somatostatin receptors in neuroblastomas(22,30,31) and medulloblastomas(32,33), as well as the clinical utility of somatostatin receptor scintigraphy(33,34) and radioreceptor-guided surgery(35).

In the current study we have identified the predominant somatostatin receptor subtypes expressed in medulloblastoma and PNET through the use of RT-PCR and quantitative autoradiography. As a control group we studied tumors of the ESFT, a group of tumors histologically similar to cPNET, but well characterized on a molecular level. In addition, we have investigated the antiproliferative effects of somatostatin in cPNET cell lines. Our study gives additional support for the theory that medulloblastoma may be derived from the EGL of the developing cerebellum.

METHODS

Culture of cPNET cell lines. Ten human cell lines were analyzed; D283 Med, D341 Med, and Daoy were obtained from the American Type Culture Collection (Rockville, MD)(36–38). Cell lines MHH-MED-1, -3, -4; Wü 1580; MEB-MED-8S; and MHH-PNET-5 were generated by one of us (T.P.)(39). CHP 707 m was the kind gift of A.E. Evans (Department Pediatric Hematology and Oncology, Children's Hospital, Philadelphia, PA)(40) and was generated from bone marrow metastases of a 33-y-old individual with a supratentorial cPNET. In contrast to the other cell lines, MHH-PNET-5 was generated from a spinal PNET and is classified as a peripheral tumor.

D283 Med, D341 Med, and Daoy were cultured in MEM, 10% heat-inactivated fetal bovine serum (FBS, GIBCO, Grand Island, NY), 0.1 mM nonessential amino acids, 100 mM penicillin-streptomycin (pen/strep), and 1 mM sodium pyruvate. The cell lines MHH-MED-1, -3, -4; Wü 1580; MEB-MED-8S; and MHH-PNET-5 were cultured in high glucose formulation Dulbecco's modified Eagle's medium (GIBCO), 10% heat-inactivated umbilical cord serum, and 4 mM L-glutamine. CHP 707 m was cultured in RPMI 1640, 15% FBS, 2 mM L-glutamine, 0.1 mM nonessential amino acids, and 100 mM pen/strep. Daoy and CHP 707 m each grew as a monolayer, D283 Med and MHH-PNET-5 grew partially adherent; all other cell lines grew in suspension. The cells were cultured in a humidified atmosphere of 95% air, 5% CO2 at 37°C. Adherent cells were harvested when in exponential growth (i.e. when 80 to 95% confluent), using a cell scraper (binding experiments) or 0.05% trypsin (GIBCO) (RT-PCR analysis). Suspension culture cells were harvested by centrifugation when between 1 × 106 and 2 × 106 cells/mL. Sterile human umbilical cord serum was prepared by centrifugation and filtration of whole coagulated blood harvested from umbilical veins immediately after birth using a protocol approved by the Institutional Review Board of Children's Hospital. All other cell culture reagents were obtained from GIBCO. Cell lines were tested for mycoplasma at 6-mo intervals by fluorescent Hoechst 33258 stain, after culture for 3 passages in antibiotic-free medium(41).

Tumor tissue procurement and preparation. Tumor tissue specimens of human PNET were obtained through the CHTN (Pediatric Division, Columbus, OH). Samples had been collected at the time of surgery and had immediately been frozen at -80°C. All available patient data, including age at diagnosis, sex, primary tumor site, tumor stage, histopathologic examination, and survival were provided by the CHTN (Tables 1 and 2) according to the guidelines established by the National Cancer Institute and the Children's Hospital Institutional Review Board. Frozen tissue sections had been examined by a neuropathologist to confirm diagnosis as well as to guarantee purity of tumor tissue and absence of necrosis.

Table 1 Clinical characteristics of patients with central PNET
Full size table
Table 2 Clinical characteristics of patients with tumors of the peripheral PNET and ESFT
Full size table

Isolation of total RNA. A modification of the acid guanidine thiocyanate-phenol-chloroform extraction method using the TRIzol reagent (GIBCO) was applied to isolate total RNA from primary tumor tissue and cell lines. Briefly, cells were washed with PBS. Adherent cells were placed in 3 mL TRIzol reagent per 25-cm2 culture flask. Cells grown in suspension were placed in 1 mL reagent per 5-10 × 106 cells. Cells were then lysed by guanidine, and the cell lysate immediately exposed to the phenol component of the reagent. This mixture was left at room temperature for 5-10 min and then either frozen at -80°C or further processed. Finally a standard phenol chloroform extraction was performed(42). A similar method was applied to frozen tumor tissue. Tissue samples were weighed into Eppendorf tubes, then kept frozen with liquid nitrogen while being ground to a powder with a mortar and pestle. Frozen sections adjacent to each sample were checked for tumor tissue by a neuropathologist in accordance with CHTN procedures. The TRIzol reagent was added to a volume of 1 mL per 50 mg of tissue. The mixture was homogenized with a Brinkman Polytron for 1 min on ice. The remaining mixture was further treated as above. Isolated total RNA was redissolved in diethylpyrocarbonate-treated, autoclaved water and quantified measuring the absorption at 260 nm with a UV spectrophotometer. All samples were treated 1-4 times with DNase (Message Clean Kit, GenHunter) to guarantee absence of DNA contamination. Total RNA was analyzed on 0.8% agarose gels; only samples confirmed for RNA purity as demonstrated by lack of extraneous ethidium bromide-stained bands were used for further analysis. Additionally gDNA contamination was excluded by use of gDNA-specific primers for TP53 in control PCR reactions. Whenever sufficient material was available experiments were performed in triplicate.

RT-PCR analysis of total RNA samples. Total RNA was analyzed using a GeneAmp RNA PCR kit (Perkin Elmer, Roche Molecular Systems, Inc., Branchburg, NJ); 200-ng aliquots of RNA were denatured for 5 min at 70°C and then cooled at 4°C. The reverse transcriptase reaction mix (25 mM MgCl2; 10× PCR buffer II [500 mM KCl, 100 mM Tris-HCl, pH 8.3], distilled, autoclaved water, 1 mM dinucleotides, 2.5U/µL MuLV reverse transcriptase, 1U/µL RNase inhibitor, and 2.5 µM random hexamers) was added together with 200 ng RNA template. The total reaction volume of 20 µL was incubated at 23°C for 10 min and 42°C for 15 min followed by 99°C for 5 min. The primers for PCR amplification were designed using the published sequences for the respective genes as shown in Table 3. Primer sequences with as small as possible homology between sst receptor subtypes were selected. PCR primers for sst5 were tested on a plasmid carrying the entire coding sequence of this receptor as a positive control. The c-abl gene, which is constitutively expressed in mammalian cells(43), was used as an additional positive control for both RNA purity and RNA quality. Lack of DNA contamination in RNA preparations was assessed by analysis of gDNA amplification using c-abl primers that cross an intron (expected (gDNA) 2016p (cDNA) products for c-abl 764 bp and 201, respectively) and additionally by gDNA-specific primers for the TP53 gene. PCR buffers were added and samples were heated to 94°C for 5 min followed by 33 PCR cycles of 1 min at 63.8°C, 1 min at 72°C, and 1 min at 95°C. The PCR reaction was finished by a prolonged extension time of 72°C for 10 min.

Table 3 PCR primers for analysis of somatostatin and its corresponding receptor subtypes
Full size table

The cDNA products of PCR were analyzed using 0.8% to 2% agarose gels; identity of PCR products was confirmed by Southern hybridization analysis(42). PCR products were transferred by capillary attraction overnight from 1% agarose gels onto nylon membranes (Hybond-N) and cross-linked by UV light (Stratalinker, Stratagene, La Jolla, CA). Nested 42-mer oligonucleotide probes were designed for the somatostatin gene and all five somatostatin receptor subtype genes and end-labeled with 32P-dCTP. Boiling for 10 min, followed by cooling on ice, denatured the labeled oligonucleotides. The label mix was added to the membrane-hybridization reaction and incubated at 42°C overnight.

All membranes were washed twice under low stringency conditions with 2× SSC/0.1% SDS at room temperature on a shaker for 15 min each and under high stringency twice with 0.1× SSC/0.5% SDS at 65°C with agitation for two times 20 min each followed by autoradiography. Results of RT-PCR and Southern hybridization were statistically evaluated by χ2 test analysis.

Receptor binding experiments. Plasma membranes were harvested from cell lines Daoy, D283 Med, MHH-PNET-5, MHH-MED-1, and four primary PNET. PNET cell lines or primary PNET tissues were harvested into buffer A (20 mM HEPES, 2 mM MgCl2, 5 mM sodium EDTA, 1 mM 2-mercaptoethanol, 150 mM NaCl, and 50 mg/mL phenylmethylsulfonyl fluoride [PMSF], at pH 7.4) at a concentration of 1 × 107 cells/mL when in exponential growth (90-95% confluence) or 50 mg tissue/mL. Membranes were prepared according to our previously published methods(22). Membranes were resuspended in buffer A and frozen at -80°C until further use. Protein concentration was measured by a commercially available BCA™ kit (Pierce, Rockford, IL) by a method modified after Lowry et al.(44). Binding experiments were performed in Buffer S (10 mM CaCl2, 50 mM HEPES, 25 mM MgCl2, 50 µg/mL bacitracin, 200 kIU/mL aprotinin, and 0.02 µg/mL PMSF) with 100- to 200-µg protein aliquots of cell membranes using SS-14 (Bachem, King of Prussia, PA) and octreotide (Novartis, East Hanover, NJ), to compete with radiolabeled 125I-SS-14 (2000 Ci/mmol; NEN, Boston, MA). Membrane binding was performed in buffer S for 60 min at 17°C in a shaking water bath. Binding to Daoy cells was performed in 24-well plates. Cells were washed with unsupplemented medium and then incubated in a total volume of 0.5 mL with the corresponding agents for 30 min at 37°C. Binding results generated from whole cells of this cell line were compared with membrane binding from the same cells. Affinity constants were estimated using LIGAND(45), and binding curves were graphed using Prism GraphPad(46). Experiments on cell lines and whenever enough tissue was available of primary tumors were performed in triplicates.

Receptor autoradiography. Twenty primary PNET were available in sufficient quantity to perform receptor autoradiography (4 medulloblastomas, 4 PNET, 8 Ewing's sarcomas, and 4 peripheral PNET). Somatostatin receptor density was measured as described previously(10). Briefly, frozen samples were cut on a cryostat in 10- and 20-µm sections and mounted on clean microscope slides that were stored at -20°C for at least 3 d. Tissue sections were incubated for 2 h at room temperature with 125I-labeled tyrosine-3 octreotide (2000 Ci/mmol). After washing, the sections were opposed to 3H-Hyperfilms (Amersham, Little Chalfont, UK) and exposed for 1-10 d. Nonspecific binding was determined in parallel sections incubated with the same concentration of labeled peptide in the presence of 10-6 M of the corresponding unlabeled peptide. In selected cases displacement experiments with successive sections of the same tumor were performed using increasing concentrations of unlabelled compound and other biologically active and inactive peptides. The resulting autoradiograms were quantified with a computer-assisted image-processing system. Labeled tissue sections were exposed together with standards (autoradiographic 125I-microscales, Amersham) that contained known amounts of isotope, cross-calibrated to tissue-equivalent ligand concentration. The image analyzer was calibrated to the standards. When the OD measured over a tissue area was at least twice that of the nonspecific binding section, a tumor was considered sst-positive(47).

[3H]Thymidine incorporation assay. The effect of somatostatin on cell proliferation was assessed by a [3H]thymidine incorporation assay(48). PNET cell lines Daoy, MHH-MED-3, Wü 1580, MEB-MED-8S, D283 Med, and D341 Med in exponential growth were washed three times with complete medium and plated at concentrations of 1 × 104 to 5 × 104 cells per well in 96-well flat-bottom microtiter plates containing various concentrations of active and inactive peptides and added in concentrations of 10-6 to 10-8 M. After a culture period of 68 h at 37°C, 5% CO2, the cells were exposed to a 4-h pulse of 0.5 mCi [3H]thymidine (25 Ci/mmol, Amersham, Braunschweig, Germany). Finally the cells were harvested on glass fiber filters, and the incorporated radioactivity was measured on a scintillation counter (Packard, Frankfurt, Germany). Assays were performed in triplicate.

A standard Student's t test was applied for calculation of the statistical differences between treated and untreated cells.

RESULTS

Patients. Tumor tissue was obtained from 29 patients with cPNET (6 PNET, 23 medulloblastomas) and 17 patients with ESFT (10 Ewing's sarcoma, 7 peripheral PNET) as shown in Table 1 and 2. These patients ranged in age from birth to 25 y at diagnosis (the minimal age in the cPNET group was 5 mo, maximal 19 y; ESFT minimal age was newborn, maximal 25 y).

In the cPNET group, eight patients were dead at onset of the study (six medulloblastoma patients, two patients with PNET). Two patients with PNET had been lost to follow-up. The male to female ratio in the cPNET group was 2.2:1 (20 male patients, 9 female). In the ESFT group five patients were alive, nine were dead, and three had been lost to follow-up. The male to female ratio was 7.5:1 (15 male, 2 female patients). We did not detect any statistically significant correlation between somatostatin or somatostatin receptor expression with clinical outcome or sex of the patients studied.

RT-PCR and Southern analysis on primary tumors and cell lines. The gene for the somatostatin peptide (SS-14) was expressed in 9 of 29 cPNET and only 1 of 17 ESFT; mRNA for SS-14 could not be detected in any of the peripheral PNET whereas 1 Ewing's sarcoma expressed SS-14 mRNA. The difference in SS-14 expression between cPNET and ESFT was statistically significant (p < 0.05 by Fisher's exact test; χ2 = 3.985). Twenty-five of 29 cPNET (86%) expressed mRNA for at least one sst (Table 4 and Fig. 1A). Two medulloblastomas and two PNET were negative for sst. In the 17 ESFT, 5 tumors (29%) did not express any mRNA for either somatostatin or any one of the five sst.

Table 4 Expression of SS-14 and its receptors in primary cPNET and peripheral PNET/Ewing's sarcoma-RT-PCR and Southern blot results
Full size table
Figure 1
figure 1

A, Southern hybridization analysis of RT-PCR products for somatostatin receptor subtype sst1 and sst2, and c-abl in 12 central PNET. cDNA Southern blots were exposed for 2 h and again for 10 h. The pictures are derived from the 10-h exposure films. Only samples positive at 10 h and on initial ethidium bromide gels were considered positive. Tumors 2, 10, 11, and 12 are supratentorial PNET; all other samples are medulloblastomas. No somatostatin receptor expression is observed in supratentorial PNET 10. Tumor 1 does not show amplification products for sst2. B, Southern hybridization analysis of RT-PCR products for somatostatin receptor subtype sst1 and sst2, and c-abl in 12 tumors of the Ewing's sarcoma family (ESFT). cDNA Southern blots were exposed for 2 h and again for 10 h. Only samples positive at 10 h and on initial ethidium bromide gels were considered positive. The pictures are derived from the 10-h exposure films. Samples 1-8 are derived from Ewing's sarcomas, and samples 9 and 10 are derived from peripheral PNET. All samples are positive for c-abl. Samples 5, 6, and 8 do not show expression of sst1 or sst2. In addition, samples 1, 2, 4, 10, and 12 were graded as negative according to our criteria.

Full size image

Of the five somatostatin receptor genes, sst1 was expressed equally in cPNET (83%) and ESFT (71%) (Table 4, Figs. 1A and 1B). The sst2 was expressed in 22 of 29 cPNET (76%) compared with only 6 of 17 of ESFT (35%). The sst2 gene was found only in two peripheral PNET (95 12 P001 and 93 02 U01), whereas four Ewing's sarcomas amplified the sst2. Through χ2 and Fisher's exact test analysis, we calculated significant differences between cPNET and ESFT for sst2 expression (76% versus 35%; p = 0.006; χ2 = 7.4).

Somatostatin receptor subtypes sst3 and sst4 were also more often expressed in cPNET than ESFT; however, sst5 expression was lacking in all 29 cPNET and all 17 peripheral PNET examined. A vector containing the full coding sequence for sst5 was used as a positive control for the PCR and Southern hybridization. All samples analyzed showed good RNA quality as demonstrated by RT-PCR of the housekeeping gene c-abl.

All of the PNET cell lines analyzed expressed mRNA for at least one sst (Table 5 and Fig. 2). Most commonly this was the sst2 (9 of 10), with sst1 being expressed by 8 of the 10 cell lines. The sst3 could only be detected in 3 of the 10 cell lines under investigation. The sst4 (1 of 10) and sst5 (0 of 10) were only rarely or not at all expressed. Somatostatin peptide mRNA was detected in 4 of 10 cell lines.

Table 5 Expression of SS-14 and its receptors in PNET cell lines tested by RT-PCR and Southern hybridization
Full size table
Figure 2
figure 2

Southern hybridization analysis of RT-PCR products for somatostatin receptor subtype sst1 and sst2, and c-abl in 10 PNET cell lines (see also Table 5). cDNA Southern blots were exposed for 10 h with double intensifying screens. The cell lines Daoy and D341 show no signal for sst1. sst2 is expressed by all cell lines except MHH-PNET-5.

Full size image

Quantitative autoradiography and membrane binding. Ten of the 20 PNET analyzed by octreotide autoradiography expressed high densities of receptor protein on their cell surface (Figs. 3 and 4 and Table 6). Although all of the cPNET exhibited moderate to very strong binding of the radioligand (Fig. 3), only two of the ESFT showed significant radiotracer uptake. One of these (95 08 P101), which was clearly diagnosed as a peripheral PNET by expression of the MIC-2-equivalent HBA-71, exhibited a very high density of somatostatin receptors. In the group of cPNET, the PNET had 4- to 25-fold lower radiotracer uptake compared with the medulloblastoma tissues (Table 6). All tissues that were positive on autoradiography expressed mRNA for at least two sst. Of the 10 ESFT that were negative on autoradiography, 5 expressed mRNA for at least one receptor subtype. This phenomenon could be explained by the different sensitivities of the assays used or by a posttranscriptional defect in the tumor cells that prevents translation into functional receptor protein(49). Each tumor that demonstrated significant octreotide binding also demonstrated mRNA for sst2; this observation confirms the preferential binding of octreotide to sst2 subtype as has been demonstrated by others(50,51). Autoradiography clearly showed expression of sst on the tumor cells.

Figure 3
figure 3

Somatostatin receptor autoradiography in two different cases of PNET. A-C, supratentorial PNET (96 02 P366); D-F, peripheral PNET (95 12 P001). A, D, Hematoxylin and eosin-stained sections. Arrowhead, tumor cells; bars indicate 1 mm. B, E, Autoradiograms showing total binding of 125I-Tyr3-octreotide. All tumor cells are labeled; however, the tumor tissue in (E) has a much higher receptor density. C, F, Autoradiograms showing nonspecific binding of 125I-Tyr3-octreotide.

Full size image
Figure 4
figure 4

Displacement curves in a medulloblastoma (upper) and in a Ewing's sarcoma (lower). In both cases, high-affinity displacement of 125I-Tyr3-octreotide by SS-14 is observed, whereas SS-28(1–12) is inactive.

Full size image
Table 6 Comparison of RT-PCR, quantitative in vitro autoradiography, and membrane binding assays for somatostatin and somatostatin receptors on 20 primary PNET
Full size table

Sufficient tissue was available from four tumors to perform competitive binding experiments with SS-14. The one medulloblastoma tested (92 07 P305) with this technique showed high-affinity binding with affinity (kd) of 1.2 nM in accordance with the RT-PCR experiment, which showed expression of the sst1-4 subtypes. Two Ewing's sarcomas (90 02 EE02, 90 04 EE01) that were negative by RT-PCR and autoradiography (Table 6) did not show any membrane binding, whereas one tumor (93 09 U07) positive for sst1-4 mRNA, but negative on autoradiography, did not exhibit any binding in the membrane preparations. This result possibly reflects the lower sensitivity of membrane binding and autoradiography compared with RT-PCR or a defect in the further translation of expressed transcripts.

Of the four cell lines investigated by membrane-binding experiments, only D283 Med bound somatostatin and its congener octreotide to a significant extent (Table 7). The kd for SS-14 was calculated at 0.98 ± 0.45 nM and that for octreotide was slightly higher at 1.9 ± 1.2 nM. This cell line expressed mRNA for sst1-4. Even though the cell lines Daoy, MHH-MED-1, and MHH-PNET-5 expressed sst2 mRNA and MHH-MED-1 expressed sst1 mRNA, these cell lines did not bind SS-14 or octreotide to a significant extent.

Table 7 Inhibition of medulloblastoma cell lines after SS-14 treatment as measured by [3H]thymidine incorporation
Full size table

[3H]Thymidine incorporation assay. Eight cell lines were tested for their response to treatment with somatostatin (Table 7). A significant reduction in [3H]thymidine incorporation in the cell lines D283 Med and D341 Med was observed in response to 10-6 M SS-14 (p < 0.005 for D283 Med and p = 0.005 for D341 Med). Both cell lines express the sst2 subtype. None of the other cell lines examined exhibited any significant response to treatment with somatostatin. A change in the differentiation pattern of the cell lines (e.g. outgrowth of neurites) after treatment with SS-14 was not observed.

DISCUSSION

In this study we present the first data on sst expression in cPNET and ESFT. Expression of mRNA for sst was identified by RT-PCR and confirmed with Southern hybridization of cDNAs, and receptor protein was demonstrated by quantitative autoradiography. These data correlate with our earlier demonstration of high-density somatostatin receptor expression in a limited number of medulloblastoma samples by autoradiography alone(33,52). We now demonstrate the sst subtype by RT-PCR, as well as the receptor protein by autoradiography, in a total of 29 cPNET and 17 ESFT. This is the first study analyzing somatostatin receptor mRNA and protein expression simultaneously in childhood brain tumors and neoplasias of the ESFT.

The somatostatin receptor subtypes most commonly expressed in medulloblastomas and cPNET are sst1 and sst2, whereas only sst1 is highly expressed in ESFT. High expression of sst2 in medulloblastoma was predicted from in vitro autoradiography and radioreceptor scintigraphy using the sst2-preferring analog, octreotide, as radioligand(33).

The sst2 expression pattern clearly distinguishes medulloblastomas from ESFT. This is evident at the mRNA level, where >75% of cPNET express sst2 compared with only 35% in the ESFT (p = 0.006). Even more important are the binding data by autoradiography, which reflect receptor protein expression. Only one of seven Ewing's sarcomas and one of five peripheral PNET showed any radiotracer uptake. In addition, there appears to be a difference in sst protein expression between medulloblastomas and PNET. Although the sst density on autoradiography in medulloblastomas ranged from 8 720 to 19 873 dpm/mg tissue, PNET displayed much lower levels of binding (between 797 and 2277 dpm/mg tissue). Admittedly our limited sample number, which is because of the rare occurrence especially of PNET, is not large enough to allow any further conclusion and needs validation by larger numbers in collaborative efforts.

Ewing's sarcomas as well as peripheral PNET are derived from cells of the neural crest. Previously we demonstrated that in neuroblastoma, another embryonal tumor derived from the neural crest, malignant cells with a higher degree of differentiation exhibited higher sst densities(31). A possible explanation for the high sst protein density shown by autoradiography in two ESFT samples is accordingly a higher degree of differentiation along the neural crest lineage. Interestingly only one peripheral PNET and one metastatic tumor of a Ewing's sarcoma (95 08 P101 and 95 12 P001) demonstrated high densities of somatostatin receptor protein on autoradiography, whereas all other ESFT tested by this technique were negative. We did not observe any difference in conventional histopathology or in the clinical behavior of these two particular tumors; both patients died after multiagent chemotherapy. Somatostatin peptide and somatostatin receptor binding are favorable prognostic factors in neuroblastoma(11,31); however this does not appear to be true in medulloblastoma. Even though somatostatin receptors were detected at high densities as evidenced by autoradiography and at high frequency in cPNET, we did not detect any significant correlation to the clinical outcome of the individual patients. The gene for the somatostatin peptide was only expressed at low levels in both the cPNET group and the ESFT group; again no correlation between clinical course and gene expression was observed. Although the result for medulloblastomas appears to be clear, larger numbers will be needed to further evaluate the prognostic significance of somatostatin receptor protein expression in all tumors studied.

Our previous analysis of sst expression in tumors of the nervous system demonstrated very high-density somatostatin receptor expression in medulloblastomas compared with astrocytomas, which exhibited intermediate receptor density, and glioblastomas, which rarely showed any somatostatin binding by autoradiography(52). An important observation in the present study is the similarity between somatostatin binding in medulloblastoma cell lines compared with primary tumors. A similar pattern of mRNA expression was observed in both PNET cell lines and primary tumors; furthermore, radioligand binding was not always present in the primary tumor or the cell lines even though expression of sst mRNA had been detected. Only one cell line, D283 Med, bound SS-14 and octreotide with high affinity. Treatment with SS-14 inhibited [3H]thymidine uptake in D283 Med and D341 Med. A similar phenomenon was observed by Fisher and colleagues in a model of pancreatic cancer(49). These authors found high expression of sst1 mRNA in nine cell lines but observed high somatostatin binding in only one of these. This further supports our rationale for performing RT-PCR and radioligand experiments in parallel. Only by using both approaches in combination can one assess the transcriptional regulation of receptor subtype expression and the presence of functional protein on the tumor cells. The reason for the lack of complete correlation between sst mRNA and high affinity SS-14 binding is unknown. One possibility is the greater sensitivity of PCR in comparison with radioligand binding. Higher numbers of receptors or a certain density of receptor proteins may be necessary for detection of receptor protein by radioligand binding, compared with the number of mRNA molecules detected by PCR, which amplifies the signal many fold.

The somatostatin receptor subtype best characterized to date is sst2, which is also the sst most frequently expressed in medulloblastomas. It has been implicated in transmitting antiproliferative signaling through several pathways(19,53) in different tumor types ranging from neuroendocrine neoplasms, such as insulinomas, to solid breast cancer and lymphoma(54). Since the discovery of a selective ligand for the sst2 (octreotide), treatment for a variety of conditions in children has become feasible(55). Additionally, techniques such as radio-receptor-targeted radiotherapy, peptide scintigraphy, and radio-receptor-guided surgery have become feasible with radiolabeled peptide ligands that recognize sst2(35,56,57). The biologic function of these sst2 receptors in medulloblastoma remains somewhat speculative. However, this finding is highly unusual as thus far only tumors with a higher state of differentiation and a more benign clinical behavior have been demonstrated to contain high numbers of somatostatin receptors. In neuroblastoma, for instance, sst2 density correlates with lower stages and lack of MYCN amplification(31,58). In contrast, medulloblastoma is a highly malignant grade IV tumor according to World Health Organization classification(3).

Recently, Robertson et al.(59) suggested a role for GH in medulloblastomas. These authors found increased heights in children with medulloblastoma. Even though highly speculative it is possible that high expression of the sst2 serves as a counterregulatory mechanism for GH or IGF-1 action. Further studies are certainly warranted to examine this hypothesis.

Another intriguing finding in our samples is the total lack of sst5 expression in cell lines and primary tumors. To exclude technical artifacts we used a cloned fragment of the coding sequence of the sst5, which always yielded a positive result on Southern blots, as a positive control; like sst2, sst5 has been implicated in antiproliferative signaling(60). Experience in our own lab indicates that tumors derived from neural crest cells do not express mRNA for sst5(61). These observations suggest that sst5 is either downregulated during development of these tumors or is not necessary for the normal proliferation in neuroepithelial stem cells.

Even though we did not measure somatostatin content in our samples by radioimmunoassay, we observed a difference in the mRNA expression for somatostatin (SS-14) between the tumors of the ESFT family and the cPNET. As mentioned earlier, we and others(13,58) had previously noted a positive correlation between outcome and somatostatin peptide content in neuroblastoma tissue. We did not observe a statistically significant effect of somatostatin peptide expression on outcome in cPNET and ESFT. The only patient in the ESFT group (96 01 P007) who expressed the somatostatin peptide is alive after multiagent chemotherapy and local radiotherapy. The patient's tumor sample expressed mRNA for sst1 but not sst2 and did not bind octreotide on autoradiography.

From a developmental point of view, our data add support to the hypothesis that the cell of origin in medulloblastoma is derived from the EGL of the developing CNS(1,62,63). Extensive investigations have shown that the somatostatin receptor expression is highest in the EGL of the cerebellar cortex during human development(18,64–67). In these studies all five layers of the cerebellar cortex expressed somatostatin receptors at 20 wk of human fetal life. At 29 wk there was a much stronger labeling of the EGL compared with the other cerebellar layers. This pattern appears to remain stable until the physiologic involution of the EGL at 8 mo. When looking at adult cerebella in the same study, the investigators found no somatostatin binding in three of four samples studied. In a very recent study, deregulation of ErbB-2 expression was observed in the developing EGL as a possible contributing factor to the initiation and progression of medulloblastomas. These authors examined normal cerebella throughout fetal life and until 20 mo of postnatal life, as well as a large number of primary medulloblastomas(68). Our results, in conjunction with previous evidence, suggest that a precursor cell in the EGL with high sst2 density is exposed to an arresting signal by somatostatin peptide. Consequently this cell, which physiologically should have migrated to deeper layers of the cerebellar cortex, may become the target of other growth factors and thus adopt a malignant phenotype. Our observations in supratentorial PNET support this concept further. Although all of the medulloblastomas studied showed high receptor density by autoradiography, the PNET had much lower receptor densities. No other studies on somatostatin receptors in child cerebellum are available to date.

Somatostatin receptors were demonstrated in the granule cell layer of adult cerebellum in two studies(18,69). Both groups found that somatostatin possesses neurotrophic and neuromodulatory activities in the human cerebellum. We conclude that there is a close relationship between development of medulloblastomas and somatostatin receptor expression in normal development of the cerebellum.

More important than being a scientific challenge for neuro-developmentally oriented scientists, medulloblastoma and other cPNET present as a serious clinical problem. High-grade disseminated tumors and recurrent tumors have a poor prognosis. We have previously demonstrated the usefulness of somatostatin in the imaging and intraoperative localization of neuroblastomas through somatostatin receptor scintigraphy and radioreceptor-guided surgery(35,57). Furthermore, we have demonstrated that somatostatin receptor scintigraphy is suitable for the detection of primary and recurrent medulloblastomas(33). In a limited study of 20 patients with medulloblastomas, we demonstrated the potential benefits of somatostatin receptor scintigraphy. One patient with a relapse showed high radiotracer uptake, and the resected tumor was highly positive on autoradiography. In another patient, somatostatin receptor scintigraphy was negative, suggesting no remaining viable tumor, whereas computed tomography and magnetic resonance imaging raised suspicion of a relapse tumor. The following second-look surgery demonstrated only scar tissue, thus confirming the scintigraphy results(33). Even though the technique of somatostatin receptor imaging is not specific for these neoplasms, it has the potential of aiding the neuroradiologists in situations when the differentiation between recurrent tumor tissue and scar tissue is difficult.

Somatostatin has been shown to inhibit the proliferation of malignant cells in vitro and in vivo(19). Direct antiproliferative effects of somatostatin include among others up-regulation of p53 and subsequent induction of apoptosis(28). Indirect effects of somatostatin include the inhibition of growth hormone (GH) and IGF-I secretion(29). The moderate antiproliferative actions of somatostatin on medulloblastoma growth in vitro deserve further investigation and possibly clinical evaluation. It is possible that the inhibitory effects on the cell lines D283 Med and D341 Med would be more enhanced in a biologic system, in which growth factors and angiogenesis are needed for the invasion and progression of these tumors.

Clinically, somatostatin analogs have long been used in several applications in pediatric patients(55). Although the SS-14 analog, octreotide, has been used successfully in the palliative treatment of pain associated with neoplasms(70), controversy exists as to whether this compound is actually neurotoxic to normal neural cells(71). Octreotide is currently in use in combination with a variety of chemotherapeutic agents in the treatment of neuroendocrine tumors as well as in other solid tumors(72,73). A trial is under way at our institution treating children with neuroblastoma with octreotide in combination with standard chemotherapeutic agents. We speculate that patients with medulloblastoma and other cPNET (especially in relapse) could benefit from a trial of octreotide as well. Recent data(74,75) also suggest that octreotide could be used as a carrier for radionuclide therapy. Again, patients with a relapsed cPNET whose tumor has been shown to express somatostatin receptors may benefit at least on a palliative basis from a similar approach. The very high receptor density of medulloblastoma represents a particularly strong argument in favor of such a therapeutic approach. Clinically a variety of somatostatin analogs are now available for this type of clinical trials(74).

In conclusion, we have identified expression of the sst2 in a major proportion of medulloblastomas and supratentorial PNET. The sst1 was highly expressed in medulloblastomas and Ewing's sarcomas, but only rarely found in peripheral and supratentorial PNET. Thus medulloblastomas, although not clearly different from PNET, are distinguished from peripheral PNET by SS-14, sst1, and sst2 expression and from Ewing's sarcomas by expression of sst2. The very high density of these receptors on medulloblastomas supports the hypothesis that this neoplasm is derived from a precursor cell in the EGL of the cerebellum and may possibly be derived from the neural crest. With the development of antibodies for the sst2, which are currently not available commercially(76,77), unambiguous identification of the cell types expressing sst2 in normal and neoplastic cerebellar tissue will be possible. Further functional studies involving up-regulation of specific sst transfections are needed to answer critical points about the cell of origin and differentiation pathways for these stem cells. Importantly, those studies suggest that high sst2 expression can regulate growth of medulloblastoma cells. Taken together, these results support the use of techniques such as somatostatin receptor scintigraphy and radioreceptor-targeted therapy in the initial diagnostic work-up, diagnostic follow-up, and treatment of children with recurrent and high-risk medulloblastoma.