Mutation and expression analysis of the putative prostate tumour-suppressor gene PTEN.

The chromosomal region 10q23-24 is frequently deleted in a number of tumour types, including prostate adenocarcinoma and glioma. A candidate tumour-suppressor gene at 10q23.3, designated PTENor MMAC1, with putative actin-binding and tyrosine phosphatase domains has recently been described. Mutations in PTEN have been identified in cell lines derived from gliomas, melanomas and prostate tumours and from a number of tumour specimens derived from glial, breast, endometrial and kidney tissue. Germline mutations in PTEN appear to be responsible for Cowden disease. We identified five PTEN mutations in 37 primary prostatic tumours analysed and found that 70% of tumours showed loss or alteration of at least one PTEN allele, supporting the evidence for PTEN involvement in prostate tumour progression. We raised antisera to a peptide from PTEN and showed that reactivity occurs in numerous small cytoplasmic organelles and that the protein is commonly expressed in a variety of cell types. Northern blot analysis revealed multiple RNA species; some arise as a result of alternative polyadenylation sites, but others may be due to alternative splicing.

allelot-vping. w e prexiousIv identified a 9-cM interval at the lOq23/24 boundarv that is deleted in most prostate tumours (Grav et al. 1995). Recentlv a candidate tumour-suppressor gene w-ith putatix e actin-binding and tyrosine phosphatase domains has been identified at 10q23.3 and designated PTEN  MMACJ (Steck et al. 1997). Mutations in PTEN hax-e been found in tumour specimens derix ed from glial. breast. endometrial and kidnev tissue (Rhei et al. 1997: Steck et al. 1997: Tashiro et al. 1997: Aanc et al. 1997 and in a number of melanoma (Guldberg et al. 1997) and prostate adenocarcinoma cell lines : Steck et al. 1997. Germline mutations in PTEA have been identified in indix iduals " ith the autosomal dominant svndromes Cow-den disease (multiple hamartoma syndrome) and Bannaxan-Zonana sy-ndrome. These disorders confer a predisposition to hamartomas at sexeral sites. including the breast and thyroid (Liaxx et al. 1997) and macrocephalx. lipomas. intestinal hamartomatous pol ps and vascular malformations (Marsh et al. 1997).
Here x e describe PTEN mutations in primary prostate tumours.
supporting exidence that PTEN may act as a tumour suppressor in the prostate. In addition. A e describe the generation of antipeptide antibodythat detects PTEN in Western blots and localizes it within

MATERIALS AND METHODS Mutation analysis
Tumours and xenous blood samples wxere obtained from men undergoing transurethral resection of the prostate. Tumour tissue was microdissected from normal tissue and tumour and blood DNA samples xxere prepared as described previously (Phillips et al. 1994). Using primers based on intron sequences. PTEN exons A ere amplified by polymerase chain reaction (PCR) from 30 ng of tumour DNA under the follou-inc conditions: an initial 95-C denaturing step of 2 min follou-ed by 30 cycles of 95 C for 30 s. 60 C for 30 s and 72 C for 30 s in a 50-jil reaction xolume. A 1-pA aliquot of product xxas then used to seed a second 15-cxcle reaction using M 13-21 tailed primers to facilitate dye-primer sequencing. After purification by passage through a Centricon-100 column (Amicon). exons were sequenced using a PRISM M13-21 dye-primer cycle sequencing system (Applied Biosy stems). individual PCR-amplified PTEN exons 1, 6 and 8, a 120-bp fragment derived from the PTEN 5' untranslated region (UTR) and a 239-bp fragment from the 3'-UTR beyond the first polyadenylation signal, all at low stringency. Primer sequences for amplification of the 5'-UTR fragment were 5'-GGTCTGAGTCGCCTGTCACC-3' and 5'-TTAAAACCGGCCCGGGTCCC-3'; primers for amplification of the 3'-UTR fragment were 5'-GACATTCGAGGAATTG-GCCGC-3' and 5'-CAAGCCCATTCTTGTTGATAGCC-3'. PCR was performed under the conditions described above. Probes were generated by subsequent reamplification of 5 ng of PCR product for 11 rounds of 940C for 1 min, 50°C for 1 min and 72°C for 1 min in the presence of 30 gCi of [x--32P]dCTP (Hirst et al, 1992).
Polyclonal antibody preparation, Western blotting and immunofluorescence A peptide of 19 residues from amino acid positions 342 to 360 (sequence KVKLYFTKTVEEPSNPEAS) was synthesized and conjugated to keyhole limpet haemocyanin (KLH) using glutaraldehyde, 2 mg of peptide to 2 mg of protein (Coligan et al, 1994). Rabbits were immunized with the conjugate, 100 lg per immunization, in complete Freund's adjuvant on the first occasion and then on five subsequent occasions in incomplete Freund's adjuvant; the immunizations were at 2-week intervals. Antibody levels were determined by enzyme-linked immunosorbent assay (ELISA) using peptide conjugated to bovine serum albumin (BSA) and the specificity confirmed by Western blotting on bacterially expressed PTEN. Western blots were carried out on total cell lysates dissolved in SDS-PAGE sample buffer. Proteins were visualized using an alkaline phosphatase-based chemiluminescence system (Tropix, Applied Biosystems) and sized using BioRad lowmolecular-weight standards. Immunfluorescence was performed with antibody that had been purified by protein A chromatography to reduce the non-specific cell-surface fluorescence found in the serum prior to immunization. FITC-conjugated sheep anti-rabbit IgG was used as a second antibody and immunofluorescent staining was visualized using a confocal microscope.

RESULTS
Thirty-seven primary prostate tumours, of which 24 had previously shown allele loss at lOq23.3 (Gray et al, 1995; ICG unpublished data), were assessed for mutations in all nine PTEN exons (Steck et al, 1997) by direct sequencing following PCR amplification. Five mutations were identified (Table 1), four of which result in a truncated protein, supporting the hypothesis that PTEN is a prostate tumour-suppressor gene. Four of the mutations cause frameshifts (two deletions, one insertion and a complex combined deletion/ duplication event). The remaining mutation is a small intronic deletion close to an intron/exon junction and which may cause aberrant splicing by reducing the length of the splice acceptor polypyrimidine tract (Shapiro and Senapathy, 1987). None of these mutation events was detected in matched blood samples and they therefore must have arisen somatically in the tumour. Four intronic variants were also detected, each being present in both tumour and blood DNA: a single-base A-*G substitution in intron A 96 bp upstream of exon 2; a 4-bp TTTG deletion in intron B 23 bp upstream of exon 3; a 5-bp ATCTT insertion in intron D 110 bp downstream of exon 4, and a T insertion also in intron D 28 bp upstream of exon 5. The allele with the 5-bp insert was found at a frequency of 39/74 in the 37 individuals studied. The remaining variants were less common, each being identified only once.
When used to probe a multiple-tissue Northern blot (Clontech), PTEN cDNA clone 264611 (Auffray et al, 1995), comprising exons 1-7 plus 478 bp of 5' untranslated DNA (IC Gray, unpublished data), hybridizes to at least five bands common to all tissues tested with varying relative band intensities in each tissue type (Figure I a). However, when a 120-bp fragment from the PTEN 5'-UTR was used to probe a similar blot, a single band of approximately 5.5 kb was identified in all tissues (Figure lb). To determine further the relationship between the different transcripts, mRNA from the lymphoblastoid cell line BRI8 (Snary et al, 1974) was hybridized with probes derived from the PTEN 5'-UTR and with individual PTEN coding exons 1, 6 and 8. The 5'-UTR probe identified the expected 5.5-kb band observed with the multiple-tissue Northern blot (Figure lci this. a probe derived from this extra 3' sequence does not hybridize to the 2.4-kb transcript. but detects the others from 2.7 to 5.5 kb ( Figure Iciii). Two peptides were used to immunize rabbits. but only the most C-terminal of the two (amino acids 342-360) produced an antibody response. The other peptide. a 22-amino-acid sequence from position 219 to 240. did not induce an anti-PTEN response. although a good anti-carrier (anti-KLH) response was given. On Westem blots the antiserum to the C-ternminal peptide bound to a protein with an apparent molecular weight of 54.8 kDa. close to the predicted molecular weight from the PTEN amino acid sequence (48.2 kDa). Several cell lines gave an identical pattem (Figure 2). Immunofluorescence of permeabilized cells suggested that the antipeptide antibody binds to a small particulate structure within the cytoplasm of the cell (Figure 3).

DISCUSSION
Although the identification of PTEN mutations in primary prostate tumours provides good evidence that PTEN is a prostate tumoursuppressor gene. the number of mutations detected is far lower than expected (5/37) given that nearly 70% of prostate tumours show loss of the 10kq23.3 region (Gray et al. 1995: IC Gray. unpublished data). There are several possible explanations. Sequencing as a method of mutation detection is unlikely to be 100% efficient. The Britsh Joumal of Cancer (1998) 78(10), 1296-1300 0 Cancer Research Campaign 1998 nature of prostate tumour gro%vth with a lack of normal tumour boundary makes it difficult to be certain about the level of normal tissue contamination of the dissected tumour. Furthermore. w-here functional loss is associated with the later stages of tumour progression. as appears to be the case here. there may be clonal subpopulations of tumour that do not camthe mutation. In addition. gross deletions spanning one or more exons and mutations in regulatonr sequences outside the coding region would have gone undetected.
Alternatively. there is the possibilitv that mutation or loss of a single PTEiN allele may be sufficient for a tumour growth advan-ta2e: our analysis showed three tumours with loss of one PTEN allele and a mutation in the second. 21 with loss of one PTEN allele but no detectable mutation in the second and t-o with one mutant allele but no detectable loss of the second. In summanr. 26 tumours of a total of 37 (70%s ) had alteration or loss of at least one copy of PTE.N. During preparation of this manuscript. a report appeared in the literature describing inactivation of both PTEN alleles in 10 of 80 primary prostate tumours studied (Caims et al. 1997). providing further evidence that PTEN is a prostate tumour suppressor gene. Furthermore. prostate cancer has been identified in association with Cowden disease (Inaaaki and Ebisuno. 1996). which has recentlv been shown to be caused by aermline PTEN mutations (Liaw et al. 1997). However. the possibility of a further tumour-suppressor gene at I10q23.3 cannot be excluded. PTENV appears to be expressed in a wide range of cell types: this is evident from the ubiquitous expression of the mRNA in all tissues examined and from the presence of the protein in cells from several different origns. The anti-PTEN antibody indicates that PTEN is found associated with small cytoplasmic particles. an obsen-ation in keepincg with data describinc the direct v-isualization of expressed PTEN protein with the Flag epitope (Li and Sun. 1997).
A complex pattern of transcripts w-as found for PTEN in all tissues tested. similar to the profiles previously reported by- Steck et al (1997). Although some of the transcript profile may be accounted for by alternati-e polvadenylation sites. other differences are also evident. suggesting alternative PTEN splicingr or cross-hybridization of PTE.N with mRNA species of distinct but related sequence. raising, the possibility that PTEN may be a member of a wider gene family. There appear to be at least two discrete major PTEV transcripts with 5' sequence differences.
Recently. an expressed PTEN pseudogene on chromosome 9 has been identified (Kim et al. 1998: Teng et al. 1998). Crosshybridization to RNA derived from this pseudogene may therefore account for some of the PTEN transcript profile.
The broad spectrum of tumour types showing PTE,N' mutations : Steck et al. 1997. coupled with apparently ubiquitous expression. suggests that PTEV has a role in the progression of a significant proportion of tumours derived from a diverse range of tissues. The identification of germline mutations in individuals suffering from Cowden disease (Liaw et al. 1997) raises the possibility that low-penetrance germline PTEN lesions may be responsible for some breast (and other) cancers previously thought to be sporadic. As four of the five mutations descnrbed here were detected in late-stage tumours showinr metastasis (Table 1). PTEN inactivation may be involved in a pathway leading to metastatic potential: a recent analy-sis of metastatic prostate cancer tissues also implicates PTE,\T involvement in metastasis (Suzuki et al. 1998). Therefore. it could pro-e to be a useful marker for monitoring prostate tumour progression and provide information that will assist in making therapeutic decisions.