Editor—The APC tumour suppressor gene contains at least 21 exons, including four exons upstream of exon 1.1 2 Alternative splicing involves at least seven exons, including the first five exons, and creates distinct splice forms of APC RNA.1-4Exon 1 contains an in frame stop codon upstream of its initiating methionine; hence only splice forms of APCthat lack exon 1 allow exons 5′ of exon 1 to be translated. Interestingly, splice forms lacking exon 1 are enriched in terminally differentiated tissues including brain,5 intimating that protein isoforms containing domains encoded by the exons 5′ to exon 1 may be important in cellular differentiation. This study evaluates neurones and glia of the rodent nervous system and asks which cell types express APC isoforms lacking exon 1.

Medulloblastomas and glioblastomas are neuroepithelial tumours derived from neuronal progenitor cells and glial cells, respectively.6 7 Both tumour types occur at increased frequency in a subset of adenomatous polyposis coli (APC) patients8; APC patients developing these tumour types have a variant of APC known as Turbot's syndrome.9 APC is an autosomal dominant disorder caused by the inactivation of one copy of the APC gene.10 11APC patients develop hundreds to thousands of adenomatous polyps and if the colon is not removed, colon carcinoma develops.8 The relative risk of brain tumour formation is 23 times greater for APC patients when compared to the general population between the ages of 0 and 29.12 This study also examines the genetic basis of neuroepithelial tumour formation and examinesAPC as a mutational target in medulloblastomas and glioblastomas. We assayed 41 sporadic glioblastomas and medulloblastomas, five cell lines derived from neuroepithelial tumour types, and one medulloblastoma from an APC patient to examine the possibility that mutations in the 5′ exons of the APC gene are associated with neuroepithelial tumour formation. The mutation cluster region in exon 15 of the APC gene was also examined.

Adult rat cerebellum and spinal cord were dissected and stored at −20°C. Embryonic rat spinal cord was dissected from embryonic day 16 (E16) rats and stored at −20°C. Rat embryonic neurones were purified from E16 dorsal root ganglia as described by Kimet al.13 Rat Schwann cells were purified from postnatal day 1 (P1) rats as described by Brockes and Raff.14 Rat astrocyte cells were derived from P1 rats as described by McCarthy and de Vellis.15

Total RNA was extracted from samples by the guanidinium thiocyanate method.16 cDNA was synthesised using random hexamers and Stratascript reverse transcriptase (Stratagene, La Jolla, CA) according to the manufacturer's instructions. The resulting cDNA was used as template in PCR reactions with an upstream primer, UP-GGAGAGAGAATGGAGGTGCTGC, derived from exon 0.3 of mouseApc, and the downstream primer, RP-CTCTCTTTCTCAAGTTCTTCTA, in exon 3 of mouseApc. UP and RP represent the universal and reverse primer, respectively, of M13 and were used for sequencing analysis. The upstream primer anneals specifically with nucleotides 154 to 175 of exon 0.3, nucleotide number 1 marking the 5′ end of mouseApc exon 0.3 (GenBank Accession No U66412). The downstream primer anneals specifically with nucleotides 401 to 422 of Apc exon 3, nucleotide number 1 being the beginning of the initiating methionine codon of mouseApc. RT-PCR products were electrophoresed through a 3% agarose gel, stained with ethidium bromide, and photographed. Intensities of RT-PCR product bands were analysed by Image Quant software (Molecular Dynamics, Sunnyvale, CA).

All cell lines were obtained from the ATCC in Rockville, MD. The human astrocytoma cell line SW1088 (ATCC HTB12) was cultured in Liebovitz's L-15 medium (Gibco BRL, Grand Island, NY) with 10% fetal bovine serum (Hyclone, Logan, UT). The human glioblastoma cell line T98G (ATCC CRL1690) and the human medulloblastoma cell line DAOY (ATCC HTB186) were cultured in Minimum Essential Medium (Gibco BRL, Grand Island, NY) supplemented with 10% fetal bovine serum, 1% non-essential amino acids (Gibco BRL, Grand Island, NY), and 1% sodium pyruvate (Gibco BRL, Grand Island, NY). The human medulloblastoma cell lines D341 (ATCC HTB187) and D283 (ATCC HTB185) were cultured in Minimum Essential Medium-alpha formulation (Sigma, St Louis, MO) supplemented with 10% fetal bovine serum.

All human tissues were obtained with Institutional Review Board (IRB) approvals from the University of Cincinnati College of Medicine and University Hospital, and Instituto Nazionale per la Ricerca sul Cancro in Genova, Italy. Fresh frozen specimens of 17 medulloblastomas were obtained from Italy and the United States, as described by Badialiet al.17 Twenty three paraffin embedded glioblastoma samples and five corresponding normal samples were obtained through the Department of Pathology and Laboratory Medicine, University of Cincinnati College of Medicine. Genomic DNA was prepared from paraffin embedded samples as described in Wright and Manow.18 A paraffin embedded sample from a medulloblastoma and accompanying normal tissue were obtained from an APC patient treated at The Ohio State University James Cancer Hospital and Research Institute.

Normal genomic DNA was a gift from the laboratory of Anil G Menon, Department of Molecular Genetics, Biochemistry and Microbiology, University of Cincinnati College of Medicine. These samples were obtained with IRB approval from patients followed by the Hypertensive Clinics of the University of Cincinnati and the Veterans Administration Hospital, as well as from the community at large. DNA was extracted from peripheral blood as published in Su et al 19 using Puregene DNA Isolation Kit (Gentra Systems Inc, Minneapolis, MN).

PCR-single strand conformational polymorphism (SSCP) analysis ofAPC nucleotide positions 1959 to 4904 (exon 15A to 15I, according to Groden et al 10) was performed using two different running conditions as described in Varesco et al 20 and Groden et al.21 Exons 1-14 and the remainder of exon 15 were not screened owing to limited tumour DNA amounts and the clustering of most sporadic APC mutations within a small section of exon 15.

PCR-SSCP analysis of exon 0.3 of the APCgene was performed using primer pairs UP-CGAGGGG TACGGGGCTAGG and RP-ATGGGGAGCG CCCTGGTCC, UP-ATCCGCTGGATGCGGACC and RP-GGCAGCACCTCCATTCTGTCT, and UP-CTGTATTGGTGCAGCCCGCCA and RP-AAGA CAGTGCGAGGGAAAACCA. PCR-SSCP analysis of exon BS was performed using primer pairs UP-GGG GAGTCTGCTGAGAAAAG and RP-GCCTTTCAAT GGGGTAGAGC, and UP-GCTCTACCCCATT GAAAGGC and RP-ACCACCACTCACGCTCTCGA. PCR-SSCP analysis of exons 0.1 and 0.2 was performed using primer pairs UP-AGATGGCGGAGGGCAAG TAG and RP-CTTCCTCACCAACAGCCAAC, UP-GTTGGCTCGATGCTGTTCCC and RP-CCGAGA ACTGAGGGTGGTAC, and UP-CCCCTATGTACG CCTCCCT and RP-CAATGCCTGCAGCCTCACTC. All primers are shown in the 5′ to 3′ direction. PCR was performed for one minute at 94°C, one minute at 64°C, and one minute at 72°C, for 40 cycles. PCR products were labelled by incorporation of 100 μCi of 6000 Ci/mmol α³²P-dCTP (NEN Research Products, Boston, MA). SSCP was performed using MDE gel solution (FMC Bioproducts, Rockland, ME) according to the manufacturer's instructions. Running conditions for SSCP gels were 4°C at 40 W for three hours at room temperature with 10% glycerol at 40 W for seven hours.

Gel purified, abnormal SSCP conformers from exon 15 were used as templates in new amplification reactions. These PCR products were sequenced with the dideoxy termination method usingTaq polymerase and fluorescently tagged M13 universal or reverse sequencing primers on the Applied Biosystems model 373A DNA sequencer. Gel purified, abnormal SSCP conformers from exons 0.3, BS, 0.1, and 0.2 were amplified by symmetrical PCR using DNA from each conformer as template. Sequences were obtained by automated sequencing as described and were confirmed by TA cloning (Invitrogen, San Diego, CA) and manual sequencing using a cycle sequencing method (Invitrogen, San Diego, CA).

Exon 15 of APC was examined in four separate overlapping sections using germline DNA from an APC patient. Forward primers, including the T7 promoter and a Kozak sequence, and reverse primers (as described in Powell et al 22) were used to amplify the DNA samples. PCR was performed for 35 cycles as follows: initial denaturation for three minutes, 95°C for 30 seconds, at or 2°C less than the predicted T_m of the primers for one minute 30 seconds, and 70°C for one minute 30 seconds. PCR products were separated by gel electrophoresis in 1% agarose (Midwest, St Louis, MO) to screen for any large deletions. The PCR samples were used as templates for in vitro transcription and translation reactions (Protein Truncation Nonradioactive Kit, Roche, Indianapolis, IN). The resulting biotinylated proteins were resolved on a 10% SDS-PAGE gel, electrotransferred to Immobilon-P membrane (Millipore, Bedford, MA), and visualised by chemiluminescence (Boehringer Mannheim Chemiluminescence Blotting Kit, Roche, Indianapolis, IN). PCR products that generated truncated proteins were TA cloned and sequenced as previously described to identify specificAPC mutations.

To identify the splice forms of APC in specific cell types of the nervous system, we performed RT-PCR on a mixed population of cells from adult rat cerebellum, adult rat spinal cord, and embryonic day 16 (E16) spinal cord. We also performed RT-PCR on cultures of rat dorsal root ganglion (DRG) neurones derived from E16 embryos from which glia and fibroblasts had been removed by antimitotic treatment, rat astrocytes derived from postnatal day 1 (P1) brain, and rat Schwann cells derived from P1 sciatic nerve. The upstream primer was placed in exon 0.3 and the downstream primer was placed in exon 3. Each sample containing a mixed population of CNS derived cells primarily produced the 321 bp product, indicating that although 0.3/1 and 0.3/2 splice forms of Apc were present, the main Apc splice form was RNA without exon 1 (fig 1). RT-PCR performed on a population enriched for neurones also primarily produced the 321 bp product, similarly indicating an enrichment for the Apc splice form lacking exon 1 (fig 1). Analysis of RT-PCR product bands using Image Quant software showed that the 321 bp band amplified from the splice form lacking exon 1 accounts for 58% or more of RT-PCR products from mixed CNS cell type populations and in a cell population enriched for neurones. In contrast, preparations of purified astrocytes and purified Schwann cells are greatly enriched for the 0.3/1 splice form ofApc and generally lack abundant expression of the 0.3/2 splice form (fig 1). Analysis of RT-PCR product bands using Image Quant software showed that the 321 bp band amplified from the splice form lacking exon 1 accounts for 24% or less of RT-PCR products derived from either purified astrocytes or purified Schwann cells. Apc splice forms were verified by direct sequencing of representative samples. Intermediate bands on the gels represent heteroduplex PCR products as previously described.5

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Rat neurones are enriched for the 0.3/2 Apc splice form. An ethidium bromide stained 3% agarose gel of RT-PCR products is shown. The upstream PCR primer anneals with Apc exon 0.3; the downstream primer anneals with Apc in exon 3. Arrows indicate the 0.3/1 and 0.3/2 RNA splice forms of Apc and the expected sizes of the amplification products. Image analysis shows that the 0.3/2 splice form exists as a higher proportion of RT-PCR products from mixed CNS cell populations and purified neurones as compared to RT-PCR products from purified astrocytes or Schwann cells. Additional bands between the two expected bands are the result of heteroduplex DNA.5

PCR-SSCP was used to evaluate sequence variations in theAPC gene in our sporadic tumour and cell line sample set. Exons 0.3, BS, 0.1, and 0.2 were examined, as well as nucleotides 1959-4904 in exon 15. This last region ofAPC is known as the “mutation cluster region” and was the only part of APCscreened in our samples owing to limited sample sizes and the inability of PCR to generate larger DNA fragments from paraffin sections. Base changes were identified in eight of 22 unaffected individual genomic DNA samples, two of 23 sporadic glioblastomas, eight of 17 sporadic medulloblastomas, and one of five cell lines tested. Table 1 lists all base pair changes observed in these samples.

Base changes affecting amino acid sequence include a heterozygous G to A transition at position 148 of exon 0.3 in cell line SW1088 from a human astrocytoma and a heterozygous 3 bp deletion in exon 15 in a sporadic medulloblastoma (table 1). The resultant amino acid changes are Cys to Tyr and a loss of a single glutamine residue, neither of which disrupts the reading frame.

Several other sequence variations were observed in unaffected subjects, CNS tumours, and one cell line that did not alter the predicted amino acid sequence of APC; these are most likely silent polymorphisms. A site of common variation exists at position 26 of exon 0.1, as seven of 22 unaffected subjects, one of 23 sporadic glioblastomas, two of 18 sporadic medulloblastomas, and one of five cell lines contained a C to G transversion at this location. DNA from two unaffected subjects contained this transversion as well as additional base changes. One of these unaffected subjects contained a heterozygous G to T base change at position 90 and the other contained a heterozygous G to A base change at position 17 of exon 0.1. Additionally, DNA from one unaffected subject contained a heterozygous C to G base change at position 50 of exon 0.1.

One of 23 sporadic glioblastomas and one of 17 sporadic medulloblastomas were heterozygous for an additional G following a string of seven Gs located at position 34 of exon 0.1. This addition of an eighth G was confirmed by independent PCR and sequencing reactions and is not the result of PCR or sequencing artefact. Examination of genomic samples from 22 unaffected subjects did not show this single nucleotide addition. Exon 0.1 does not contain a putative translation start, hence the effect of amino acid sequence is difficult to predict.

Three samples contained either a heterozygous or homozygous T to G transition in exon 0.2 of APC. For medulloblastoma 1837, it is not possible to distinguish between germline homozygosity for this polymorphism versus a loss of heterozygosity in the tumour because normal tissue was not available. This T to G transition is located upstream from the putative translation start site within exon 0.2 and most likely would not affect APC amino acid sequence. Three medulloblastoma samples contained a heterozygous T to A transversion at position 4326 in exon 15 within the “mutation cluster region” of the APCgene. This transversion does not change the amino acids encoded within this region.

Two heterozygous base changes were observed in a medulloblastoma from an APC patient that would truncate the APC protein (table 1). The germline APC mutation was a heterozygous GTGA deletion at nucleotide 3188, resulting in a premature stop codon at nucleotide 3373. This deletion was identified using the protein truncation test and was verified by direct sequencing. Analysis of DNA from a paraffin section of the patient's medulloblastoma showed that the second APC allele carried a deletion of an adenine at nucleotide 3187, resulting in a premature stop codon at nucleotide 3373. This somatic mutation was identified by direct sequencing. TA cloning and direct sequencing were performed to verify that the two mutations occurred on separate alleles.

The APC protein is a 310 kDa protein that homodimerises,23 24 binds β-catenin,25 26axin,27 tubulin,28 29 EB1,30and the human homologue of Drosophila disks large protein.31 APC is phosphorylated by GSK3β32 and the cyclin dependent kinase CDK1/p34^cdc2.33 The current model of APC function highlights the ability of APC to down regulate β-catenin, modulate Wnt signalling pathways, and effect gene transcription.34

The homodimerisation domain of APC is located at the amino-terminus and is predicted to play a role in APC function.23 24 Exon 1 encodes the first heptad repeat of APC, which is sufficient for homodimerisation. Four APC exons have been identified 5′ of exon 1 and are alternatively included in differentAPC splice forms. Exon 1 contains an in frame stop codon upstream of its start codon, arguing that these 5′ exons, 0.3, BS, 0.1, and 0.2, are translated into APC protein products only if exon 1 is removed from the transcript. Such alternative splicing occurs in human brain frontal lobe tissue; the 0.3/2 splice form is enriched in skeletal muscle, heart muscle, cerebrum, and cerebellum of mouse.5 At least some transcripts without exon 1 encode APC isoforms with distinct domains and distinct cellular functions, as an isoform without amino acids encoded by exon 1 does not dimerise with an APC isoform containing amino acids encoded by exon 1.35 We report here that rat neurones are enriched for the 0.3/2 splice form of Apc lacking exon 1, which may have functional significance in the terminal differentiation of neuronal cells. Mixed populations of cells expressed both the 0.3/1 and 0.3/2 splice forms, but purified astrocytes and purified Schwann cells contained primarily the 0.3/1 splice form ofApc, suggesting that the cell type contributing to the abundance of the 0.3/2 splice form in the mixed population of CNS derived cells is either neuronal or oligodendroglial. Neurones were confirmed as the predominant cell type contributing to the abundance of the 0.3/2 splice form via the predominance of the 0.3/2 splice form in purified embryonic rat neurones (fig 1). Purified oligodendrocytes were not assayed for the expression ofApc splice forms. These findings are particularly interesting given the existence of a brain specific homologue of APC, APCL, reinforcing the possible importance of APC isoforms or homologues in terminal differentiation.36APCL, like the isoforms of APC without amino acids encoded by exon 1, interacts with β catenin and may therefore also mediate Wnt signalling pathways.35 36 APCL contains a novel domain at its C-terminus that interacts with 53BP2, a protein that interacts with both p53 and Bcl-2, and EB3, an EB1 homologue expressed preferentially in brain.37 38 Apc splice forms without exon 1 and containing exons 5′ of exon 1 may encode isoforms that also interact with distinct cellular proteins and, without a dimerisation domain, that allow these isoforms to play distinct functional roles in specific cell types.

APC patients are at an increased risk for developing neuroepithelial tumours such as medulloblastoma and glioblastoma.12Hamilton et al 12 first reported an APC patient whose medulloblastoma contained a homozygous protein truncating mutation. The increased risk of central nervous system (CNS) tumours in APC patients, the high expression of APC in developing and adult rodent CNS,39-42 and scattered mutational reports12 43 suggest that a loss of APC function can be a causative factor in the development of neuroepithelial tumours. This also provides a rationale for assaying sporadic neuroepithelial tumours for APC mutations. Moriet al,44 Yonget al,45 and Vortmeyeret al 46 assayed 91 sporadic neuroepithelial tumours and a cumulative total of 22 sporadic medulloblastomas respectively for mutations in theAPC gene. No alterations were found by the loss of heterozygosity assays used in these studies. The exons 5′ of exon 1 were not surveyed by any of these groups. In this study, we report the absence of frameshift mutations or premature stop codons using more sensitive techniques and assaying different exons than in previous reports.

The high frequency of sequence variation identified in exon 0.1 by SSCP and sequence analysis, and the lack of an AUG start codon in this exon, argue that exon 0.1 may not be translated in vivo. Sequence variations were found in genomic DNA from eight of 22 unaffected subjects as well as in two of 23 sporadic glioblastoma samples, three of 18 medulloblastoma samples, and one of five cell lines. One sporadic glioblastoma sample and one sporadic medulloblastoma sample contained a heterozygous addition of a base that was not observed in any of the unaffected genomic DNA samples. While this addition of a nucleotide is specific to CNS tumour samples in our sample set, it is difficult to assign a significance to this addition of a base and CNS tumour formation owing to small sample sizes. The absence of a putative translation start site in exon 0.1 also suggests that this polymorphism may not affect protein sequence.

One sporadic medulloblastoma contained a heterozygous T to G transition in exon 0.2 of APC; one sporadic medulloblastoma and the astrocytoma cell line SW1088 contained a homozygous T to G transition at the same site. This base change is located upstream from the putative translation start site within exon 0.2 and most likely does not affect amino acid composition. This base change is most likely a silent polymorphism.

A heterozygous, silent T to A transversion was observed at nucleotide 4326 of APC in three of 17 medulloblastomas. This variant has been reported by Mandl et al 47 in one of 202 polyposis families. DNA obtained from blood from one of the patients with this silent T to A transversion contained the same nucleotide alteration, indicating a germline alteration. We were unable to obtain sample material for additional DNA extraction to confirm the presence of this rareAPC variant in the other two medulloblastomas containing this T to A transversion. One of the three subjects carrying this heterozygous base change was treated in Italy and two were treated at the Mayo Clinic; they are unrelated.

We observed two heterozygous sequence changes in this study that subtly alter the amino acid composition of APC. Neither of the base changes results in a truncation or other large scale disruption of APC amino acid sequence. The first is in the astrocytoma cell line SW1088, in which a heterozygous G to A substitution in exon 0.3 results in the single amino acid change of cysteine to tyrosine. It is unknown whether this base change and the resultant amino acid change exist in the primary tumour that gave rise to the cell line, or if the change occurred during or after the creation of the SW1088 cell line. The amino acid change is not predicted to change the secondary structure, as the region encoded by exon 0.3 is predicted to form a non-helical, non-coiled coil structure classified as an extended or loop structure in both wild type and variant APC 0.3/2 isoforms.48-50

The second base change affecting amino acid composition was identified in a sporadic medulloblastoma. A heterozygous 3 bp deletion at nucleotide 3460 results in the loss of a single glutamic acid residue at amino acid 1154. This glutamic acid is located immediately before the third of three 15 amino acid repeat motifs that are involved in β-catenin binding.25 26 The loss of a glutamic acid residue at amino acid position 1154 does not affect the sequence of the 15 amino acid motif; however, it is possible that a loss of a residue at this location disrupts the spacing between the second and third 15 amino acid repeat. These amino acid repeats are separated by only four amino acids, the fourth of which is the glutamic acid deleted in this sporadic medulloblastoma. Although this is a slight change in amino acid sequence, it is possible that the integrity of the β catenin binding region of APC is disrupted in this medulloblastoma. We were unable to obtain non-tumour tissue from this patient to establish whether this is a germline or sporadic change.

Although we found no evidence for severe disruption ofAPC in sporadic medulloblastomas, sporadic glioblastomas, or CNS derived cell lines, we did identify a disruption of both alleles of APC in a medulloblastoma from an APC patient. The germline APCmutation was a GTGA deletion at nucleotide 3188 that prematurely terminates APC. Analysis of DNA from a paraffin section of the patient's medulloblastoma showed that the secondAPC allele carried a deletion of an adenine at nucleotide 3187 that also terminates APC. Both alleles encode a protein consisting of 1062 normal amino acids followed by 61 novel amino acids and a stop codon. The protein truncation occurs between the first and second 15 amino acid repeats that are involved in β-catenin binding.26 The sporadic mutation in the medulloblastoma is most likely not the result of somatic recombination or isodisomy, as the single base deletion in the second allele of the medulloblastoma is different from the germline deletion.

The possibility exists that additional base changes inAPC could have been identified using more sensitive techniques than SSCP analysis or by more extensive analyses. Indeed, a recent study showed missense mutations inAPC in two of 46 sporadic medulloblastomas.51 Additionally, the possibility exists that mutations in sporadic medulloblastomas and glioblastomas occur in different gene targets that affect the same signalling pathway as APC. Examples include Huang et al 51and Zurawel et al,52 where oncogenic β-catenin mutations were found in three of 67 and four of 46 sporadic medulloblastomas respectively. Activated, or oncogenic, β-catenin is not down regulated by APC and results in non-regulated WNT signalling through β- catenin.53 Screening of our DNA set for sequence alterations in exon 3 of β-catenin by SSCP analysis and direct sequencing did not show any nucleotide changes (data not shown). Therefore, although specific splice forms ofAPC are highly enriched in neurones, these splice forms do not appear to be mutational targets in neuroepithelial tumour formation.