Introduction

PTEN (phosphatase and tensin homolog deleted on chromosome 10) was identified by positional cloning as a candidate tumor suppressor gene located at chromosome 10q23 that was mutated in multiple types of advanced cancers, hence its other designation as MMAC (Li et al., 1997a; Steck et al., 1997), and as a novel protein tyrosine phosphatase (TEP1) through database homology searching (Li and Sun, 1997). Mutational analyses of a wide range of spontaneously arising tumors have revealed mutations of PTEN in glioblastoma, endometrial cancer, metastatic prostate cancer, malignant melanoma, mammary adenocarcinoma, and various other tumor types (Cantley and Neel, 1999). Heterozygous mutations of PTEN have also been identified in autosomal dominant hamartomatous syndromes such as Cowden disease (Cantley and Neel, 1999; Liaw et al., 1997; Marsh et al., 1997, 1999). These hamartomatous syndromes are also associated with the development of various types of malignancies, for example, individuals with Cowden disease have an increased incidence of breast and non-medullary thyroid cancer.

There is considerable experimental evidence showing that PTEN acts as a tumor suppressor (Cantley and Neel, 1999; Di Cristofano and Pandolfi, 2000; Stambolic et al., 1998). Owing to the ability of PTEN to dephosphorylate the D3 position of PI(3,4,5)P3 (Maehama and Dixon, 1998; Myers et al., 1998), introduction of PTEN into cell lines lacking this phosphatase inhibits cell growth and anti-apoptotic signaling pathways that lie downstream of phosphatidylinositol 3-kinase (PI-3K). Consistent with its role as a tumor suppressor, targeted disruption of the murine Pten gene results in an accelerated rate of tumorigenesis in Pten+/− mice (Di Cristofano et al., 1998; Podsypanina et al., 1999), with neoplasms that have been examined exhibiting spontaneous mutations of the second, untargeted allele (Suzuki et al., 1998). These in vivo results have suggested that loss of Pten phosphatase activity is penetrant with respect to tumor development.

Except for the report that PTEN is downregulated by TGF-beta (Li and Sun, 1997), and evidence of PTEN promoter silencing as a functionally important epigenetic event in subsets of tumors (Mutter et al., 2000; Zhou et al., 2000), little is known about the physiological regulation of PTEN expression. Furthermore, there have been no reports characterizing the structural elements or functional regulation of the isolated PTEN gene promoter. Although PTEN activity may well be determined by post-translational modifications (Georgescu et al., 1999; Vazquez et al., 2000), PTEN gene transcription levels may also be an important determinant of PTEN activity within normal cells. PTEN protein levels may be closely regulated, as experimentally induced alterations in the levels of this phosphatase appear to have effects on cell growth and survival (Di Cristofano and Pandolfi, 2000). For example, Pten heterozygous mice develop a spontaneous non-neoplastic lymphoproliferative/autoimmune disease (Di Cristofano et al., 1999; Podsypanina et al., 1999), implying that a ∼50% reduction in Pten protein level is capable of deregulating the murine immune system over time. Thus the precise identification of evolutionarily conserved cis-acting control elements within the PTEN promoter will be critical to understanding how PTEN activity is normally regulated in vivo.

Loss of heterozygosity (LOH) within and proximal to the PTEN locus is frequently found in cancers. However, LOH does not appear to be invariably accompanied by PTEN mutations in tumors such as meningiomas, some glioblastomas, primary prostate tumors, and sporadic breast carcinomas, raising the possibility that other as yet unidentified tumor suppressor genes lie within the 10q region (Bostrom et al., 1998 ; Feilotter et al., 1998, 1999; Fujisawa et al., 1999; Simpkins et al., 1998). Although 220 kb of the human PTEN genomic locus has been characterized (human BAC 265N13), the organization and orientation of other genes in the vicinity of this locus remain to be fully elucidated.

The pufferfish, Fugu rubripes, has been proposed as a model vertebrate genome because of its compact size of 400 Mb (Brenner et al., 1993). Its genome is devoid of dispersed repetitive elements which constitute a significant fraction of the non-coding sequence in the human genome. Thus the intergenic regions in the Fugu genome are relatively short and considerably less complex than their mammalian counterparts. Fugu is therefore an attractive model for analysing genomic organization and promoter sequences. As a first step towards further defining the human PTEN gene locus and promoter structure, we have sequenced the Fugu pten (fpten) gene locus. This sequence (56 kb) has allowed us to assemble human BAC clones to provide the order and orientation of three human genes, hPTEN, hMINPP, and hPAPSS2. Furthermore, we show that the ∼2 kb fpten promoter is active in human cells, thus constituting a useful tool to identify conserved regulatory elements. Lastly, we find that the fpten phosphatase acts as a growth and tumor suppressor in human glioblastoma cells, demonstrating functional conservation with mammalian PTEN.

Results and discussion

Fugu pten locus

The Fugu pten locus1 (Figure 1a) contains four complete genes that show homology to known human genes. They are the multiple inositol polyphosphate phosphatase (fminpp1) gene, 3′-phosphoadenosine 5′-phosphosulfate synthase 2 (fpapss2) gene, and the omega class glutathione-S-transferase (fgsto) gene. The last three exons of another fminpp gene (designated fminpp2) are present at the 5′ end of the locus. The identical exon-intron organization and the high homology in the coding sequences (82% aa identity) between the two fminpp genes suggest that they are the result of a tandem duplication of an ancestral fminpp gene. In addition to these known genes, we identified three hypothetical genes (h1, h2 and h3) which were predicted by both GRAIL and Genscan. Hypothetical gene h2 is a single exon gene that codes for a protein of 534 amino acids. Genes h1 and h3 comprise multiple exons and code for proteins of 230 and 599 amino acids respectively. No major repetitive sequences are found in this locus.

Figure 1
figure 1

Schematic diagram of the Fugu and human PTEN loci. (a) The Fugu sequence was obtained from cosmids c141O7, c62M14 and c96J15. Human PTEN and PAPSS2 data are from GenBank (accession number AF067844) and Kurima et al. (1999), respectively. Arrows represent genes and indicate the direction of transcription. Exon-intron structure is indicated by the open boxes and thin line. Note the different scales for the Fugu and human loci. (b) Representation of the deduced ordering of human BAC clones containing sequences for hPTEN (RP11-165M8, 265N13), hMINPP1 (RP11-57C13), and hPAPSS2 (RP11-57C13, RP1177F13)

Full size image

The exon-intron structures of the Fugu genes, minpp1, papss2 and pten, are identical to those of their human homologs (Chi et al., 1999; Kurima et al., 1999; GenBank Accession number AF067844). However, the Fugu genes are highly compressed compared to the human genes. The Fugu minpp1, papss2 and pten genes span 3.4 kb, 6.5 kb and 7.3 kb respectively, whereas their human homologs occupy >22 kb (Chi et al., 1999), >85 kb (Kurima et al., 1999) and 102 kb (BAC 265N13, GenBank Accession number AF067844), respectively. Thus, the Fugu genes are seven to 14 times smaller than the human genes. The difference in the sizes between the Fugu and human genes are mainly due to the enormous introns found in the human genes (Table 1). Fugu pten encodes a protein with 89% amino acid identity to its human counterpart (Figure 2). The Fugu genes minpp1, papss2 and gsto encode proteins with 38, 79, and 55% amino acid identity, respectively, to the human proteins (Figures 3, 4 and 5).

Table 1 Intron sizes of Fugu and human PTEN and PAPSS2 genes
Full size table
Figure 2
figure 2

Alignment of Fugu, human and Xenopus PTEN sequences. Alignment was generated using the CLUSTALW program (Thompson et al., 1994). Arrow heads indicate the position of introns in the Fugu and human PTEN genes. Human, PTEN, GenBank accession number AAD38372; Xenopus pten, GenBank accession number AAD46165

Full size image
Figure 3
figure 3

Alignment of the Fugu and human PAPSS2 proteins. The intron positions (arrow heads) are conserved between the Fugu and human PAPSS2 genes. The Fugu papss2 protein is 79% identical to human PAPSS2 (GenBank accession number AF074331)

Full size image
Figure 4
figure 4

Alignment of the Fugu and human MINPP1 proteins. The arrow heads indicate intron positions in the Fugu minpp1 gene. The Fugu minpp1 is 38% identical to human MINPP1 (GenBank accession number AAD09751)

Full size image
Figure 5
figure 5

Alignment of the Fugu and human GSTO proteins. The arrow heads indicate intron positions in the Fugu gsto gene. the Fugu gsto is 55% identical to the human GSTO (GenBank accession number P78417)

Full size image

The human genes for MINPP, PAPSS2 and PTEN have all been mapped to the long arm of chromosome 10 at 10q23. Their exact position and orientation are not known, although MINPP1 is estimated to lie within 1 Mb upstream of PTEN (Chi et al., 1999). We used the Fugu pten locus sequence as a guideline to determine the order of the human genes. We first did a homology search of the non-redundant nucleotide database and the human genome draft sequence (‘High-Throughput Genome Sequences') database of the GenBank and identified several BAC clones that contain sequences for the three human genes. The complete sequence of only one of these BACs (265N13) is known, and the rest are in the ‘draft' form submitted to GenBank as ‘unordered pieces’. BAC 265N13 (218 kb) contains the complete sequence for the hPTEN gene together with 22 kb upstream and 95 kb downstream sequences. No other known gene is found on this BAC. Among other BACs, BAC RP11-57C13 (168 kb) contains sequences for all the exons for hMINPP1 and the first exon for hPAPSS2, indicating the proximity of these two genes. Another BAC, RP11-77F13 (153 kb), contains most of the exons for hPAPSS2. This BAC overlaps with BAC RP11-165M8 (169 kb) by about 23 kb and the latter contains the first five exons for the hPTEN gene. From these data we deduce that the human genes for MINPP1, PAPSS2 and PTEN lie within a region of about 500 kb, and are organized in the same orientation as their Fugu homologs (Figure 1b). We also identified three BACs (RP11-475M16, 373N18 and 99N20) that contain sequences for the human GSTO gene. These three BACs also map to chromosome 10. However, none of them overlaps with the BACs we identified for hMINPP1, hPAPSS2 and hPTEN, and therefore the location and orientation of the human GSTO gene in relation to the other three genes are unclear.

Regarding the possibility that tumor suppressor genes other than PTEN, but located in the same chromosomal region, are involved in some human cancers (Bostrom et al., 1998; Feilotter et al., 1998, 1999; Fujisawa et al., 1999; Simpkins et al., 1998), it is noteworthy that hPAPSS2 has a 20-fold higher expression in a non-metastatic human colon carcinoma cell line than in an isogenetic metastatic cell line (Franzon et al., 1999). PAPSS2 is a disease-associated gene, with mutations causing cartilage and skeletal defects in mice (brachymorphism) and humans (spondyloepimetaphyseal dysplasia) (Kurima et al., 1998; ul Haque et al., 1998) that are likely due in part to reduced sulfation of chondrocyte extracellular matrix molecules. In view of the role of PTEN as a tumor suppressor being linked to its activity as a phosphatidylinositol phosphatase, it is intriguing that the neighboring hMINPP1 gene also encodes a phosphatase with specificity for phosphoinositides. The endoplasmic reticulum-localized MINPP1 acts on inositol polyphosphates containing four or more phosphates such as InsP5 (Ins(1,3,4,5,6)P5) and InsP6, and prefers to hydrolyze phosphate at the 3 and 6 positions (Chi et al., 2000; Craxton et al., 1997; Nogimori et al., 1991). While the preferred substrates of PTEN are PtdIns(3,4,5)P3 and PtdIns(3,4)P2, it less optimally hydrolyzes Ins(1,3,4,5)P4 (Maehama and Dixon, 1998), an in vitro substrate in common with MINPP1 (Chi et al., 1999). The substrates of MINPP1 serve as binding sites for specific pleckstrin homology domains, are involved in several cell signaling pathways, and their hydrolysis has been associated with differentiation (Chi et al., 1999), indicating that mutations in this phosphatase could have profound effects on cells. It is also interesting that, like PAPSS2, MINPP1 is associated with chondrocyte function as it is highly expressed in differentiating chondrocytes (Reynolds et al., 1996; Romano et al., 1998). Nevertheless, although cellular InsP5 and InsP6 levels are elevated in Minpp1-null mice, the animals lack obvious defects and have normal chondrocyte and bone development (Chi et al., 2000). Also no germline mutations of MINPP1 were detected in patients with hamartomatous syndromes and lacking germline mutations of PTEN (Dahia et al., 2000).

The vertebrate lineages that led to Fugu and humans diverged about 400 million years ago and thus the extent to which gene order and synteny has been conserved between the Fugu and human genomes is still poorly understood. Sequencing of several homologous regions in the Fugu and human genomes, for example, has indicated that while the gene linkage has been totally conserved in some loci, in other loci synteny but not gene order has been conserved, and in yet other loci, synteny has been disrupted (Venkatesh et al., 2000). In the PTEN locus, the gene order of three of the genes are totally conserved between the Fugu and human genomes and a fourth gene found in the Fugu locus is found on the same human chromosome as the other three genes. Furthermore, the three Fugu genes are located within a short stretch of 22 kb whereas their human homologs are spread over a region of about 500 kb, supporting the view that the expansion in the human genome has occurred predominantly within the noncoding sequences (Brenner et al., 1993; Venkatesh et al., 2000).

Fugu Pten expression

Fugu fpten transcripts were detected in all the adult tissues analysed in this study (Figure 6), indicating that it expresses ubiquitously like human PTEN (Li and Sun, 1997; Steck et al., 1997). The fpten promoter spans only 2.1 kb (from the polyadenylation signal of fpapss2 to the transcription start site of fpten) and would be predicted to contain all the key elements involved in the regulation of fpten expression. We have identified putative binding sites for a variety of transcription factors, such as GATA, v-Myb, CREB, AML-1, Nkx-2 and CdxA (based on a search of the TRANSFAC database; Heinemeyer et al., 1998) within this promoter. The boundaries of the human PTEN promoter is not known as no gene is found up to 22 kb upstream of this gene. A comparison of the fpten upstream promoter region and 10 kb of the proximal hPTEN promoter region identified several 10–11 bp long elements conserved between the two promoters. While some of these elements correspond to putative binding sites for Nkx-2, AML1 and CdxA, the significance of others is unknown.

Figure 6
figure 6

Northern blot analysis of the Fugu pten expression pattern. Total RNA (10 μg each from brain and skin; 20 μg each from heart and ovary; and 40 μg each from gills, intestine, kidney, liver, muscle and testis) was probed with a full length Fugu pten cDNA clone (top panel). A Fugu actin probe was used to check the quality and quantity of RNA (bottom panel)

Full size image

To determine whether the fpten promoter is functional in human cells, the activity of the 2.4 kb sequence flanked by the translation stop codon of the upstream fpapss gene and the translation start codon of fpten was evaluated in the human glioblastoma cell line U87-MG. In the sense orientation, the 2.4 kb sequence drove expression of a luciferase reporter gene, whereas it had negligible promoter activity in the reverse orientation (Figure 7). The inclusion of an SV-40 enhancer in the plasmid further increased promoter activity 5–6-fold, suggesting that the 2.4 kb sequence contains a basic promoter which may be naturally regulated by other remote cis-elements. Thus, the DNA upstream of fpten contains promoter elements which are functionally recognized by the transcription components of a human cell.

Figure 7
figure 7

Fugu pten promoter activity in human cells. U87-MG cells were co-transfected with pCMV–CAT and a plasmid containing the Fugu pten promoter sequence in the sense or antisense (anti) orientation upstream to a luciferase reporter gene, and with or without the SV40 enhancer as indicated. Cell lysates were assayed for luciferase and chloramphenicol (CAT) activity. Fugu pten promoter activity is presented as arbitrary units representing units of luciferase activity per unit CAT activity

Full size image

Fugu pten protein

The Fugu pten gene codes for a protein of 412 residues which is 89% and 82% identical to the human and Xenopus PTEN, respectively (Figure 2). As a group, the vertebrate PTEN differ considerably from the PTEN orthologs identified in invertebrates such as Drosophila (Huang et al., 1999), C. elegans (Rouault et al., 1999) and S. cerevisiae (Li et al., 1997b). Vertebrate PTEN sequences show only 15–35% identity to their invertebrate counterparts. In fpten, the N-terminal phosphatase domain is highly conserved, with an active site and TI loop identical to human PTEN. The putative general acid Asp92 is present, as well as a residue which may bind the D5 phosphate of PI(3,4,5)P3, His93 (Lee et al., 1999). The C2-like phospholipid binding domain is also well conserved, and possesses the CBR3 loop and basic cα2 helix that in hPTEN may mediate phospholipid membrane binding (Lee et al., 1999). A portion of the C2-like domain spanning residues 286–320 is not conserved. Here the fpten has a Gln/Pro-rich region distinct from hPTEN and an eight amino acid insert lacking in both human and Xenopus PTEN (Figure 2). The comparable region of hPTEN (aa 286–311) is an unstructured or loosely folded protease-sensitive loop (Lee et al., 1999). Residues which are hot spots for tumor-derived hPTEN mutations are identical between wild-type human and Fugu genes.

Growth suppression by fpten in human cells

To test whether fpten is functionally equivalent to human PTEN, we examined its properties as a growth suppressor of the human glioma U87-MG cell line. These cells are deficient in hPTEN, and introduction of exogenous wild-type hPTEN reduces the tendency of these cells to grow in clusters lacking cell–cell contact inhibition, and suppresses proliferation and anchorage-independent growth (Furnari et al., 1997; Georgescu et al., 1999). Stable U87-MG cell lines expressing GFP-fpten were generated. Compared to the parent cells or to those transfected with empty plasmid, which grew in clusters of multicellular layers (Figure 8a,b), the fpten expressing cells grew predominantly as a monolayer up to complete confluence (Figure 8c). This was reflected by the much lower saturation density attained by the fpten-expressing cells of about one-third that of the parent U87-MG line (Figure 8d,e). The fpten cells also exhibited a slower growth rate (Figure 8d) with an average doubling time of 35 h, in contrast to a doubling time of about 27 h for the parent and vector transfected cells. Consistent with a transformed phenotype, both the parent and vector transfected cells formed large colonies in soft agar (Figure 8f,g), but fpten expressing cells formed much smaller colonies (Figure 8h). Together, these results indicate that fpten functions as a growth and tumor suppressor in human cells in a manner similar to hPTEN.

Figure 8
figure 8

Fugu pten has cell growth and tumor suppressor activities. Parent U87-MG cells (a,f), a cell line derived following transfection with empty plasmid and selection (U87-neg) (b,g), and stable lines of U87-MG cells expressing similar levels of fpten (U87-7, -8, -9 [c,h]) were examined for their phenotypes (a–c), growth (d), and saturation densities (e) attained in liquid culture, and for their abilities to form colonies in soft agar (f–h)

Full size image

Summary

The compact genome of the Fugu, with its short promoters devoid of repetitive elements, is a useful tool for completing the human genome ‘draft’ sequence and annotating it. We have indeed demonstrated the utility of the Fugu sequence for determining the order of the human BAC clones in the PTEN locus which should help in rapidly obtaining the contiguous sequence of the human genome from this locus. The observation that the short ∼2 kb Fugu pten promoter is operative in human cells, and the further definition of the operative elements, will allow comparison with the long upstream sequence of hPTEN to identify and analyse the homologous human cis-acting regulatory elements. The finding that the fpten protein shows functional conservation in human cells with hPTEN, supports the notion that the regulation of PTEN expression could also be conserved in the vertebrate lineage. Identification of the compressed pten locus in Fugu also provides the opportunity to examine the theory that other tumor suppressor genes may be located near PTEN. For example, the comparative abilities of fpten cDNA under control of the Fugu pten promoter versus that of the larger region containing Fugu pten and adjacent coding sequences to rescue Pten/PTEN-deficient mice or human tumor cell lines, or other cell lines with LOH in the region of chromosome 10q23 but lacking hPTEN mutations, can now be tested.

Materials and methods

Cloning and sequencing of Fugu pten locus

A fragment of the Fugu pten gene was amplified from the genomic DNA by PCR using degenerate primers complementary to the fourth (forward primer: 5′-CAY TAY GAY CAN GCN AAR TT-3′) and sixth exons (reverse primer: 5′-GTY TCR AAC ATC ATY TTR TG-3′) of the known vertebrate PTEN sequences and cloned into pBluescript. This fragment was used as a probe to screen a gridded Fugu cosmid library (G Elgar, UK-HGMP Resource Center) and nine positive cosmids were isolated. Three of the overlapping cosmids (96J15, 62M14 and 141O07) were selected for subcloning and sequencing. A total of 55.9 kb contiguous sequence was obtained from these cosmids (Figure 1) by a combination of shotgun sequencing and primer walking on an ABI 377 DNA sequencer. Coding sequences of known genes were identified by their homology to genes in other vertebrates, and novel genes were predicted by using the gene-prediction programs GRAIL and Genscan. The coding sequence and the transcription start site of the Fugu pten gene were determined by sequencing cDNA clones generated by RT–PCR and 5′RACE (SMART RACE cDNA Amplification Kit, Clontech).

Northern analysis

Total RNA from various Fugu tissues was fractionated on a 1.2% agarose gel containing formaldehyde, transferred to a Hybond-N nylon membrane (Amersham), and probed with α-32P-labeled Fugu pten cDNA.

Cloning and expression of Fugu pten cDNA

Forward and reverse primers (respectively 5′-TCACGAATTCATGGCTGCTATTATAAAAGAAAATGG-3′ with an added EcoRI site and 5′-AACAGGATCCTCACACTTTAGTGATTTCGC-3′ with an added BamHI site) that flank the coding sequence of Fugu pten were used for PCR with a cDNA template prepared by reverse transcription of Fugu brain RNA. Complete sequencing of the PCR product confirmed its identity as fpten cDNA and excluded the presence of any mutations introduced by PCR. The PCR fragment was cloned into pEGFP-C2 (Clontech) so as to allow expression of fpten tagged at the N-terminus with GFP.

Assay of Fugu pten promoter activity

The intergenic sequence from the translation stop codon of the upstream gene to the translation start codon of fpten was amplified by PCR using the primers 5′-AACTGATATCTTGGCACGGGAGTCAGCCAG-3′ and 5′-CATCGATATCTGTAGCAGGTGACAGGAGTC-3′. The resulting 2.4 kb fragment was inserted in the sense or antisense orientation 5′ to a firefly luciferase reporter gene in plasmids possessing (pGL3-Enhancer, Promega) or lacking (pGL3-Basic, Promega) the SV40 enhancer element. One μg of each plasmid was co-transfected with the same amount of pCMV-CAT plasmid into U87-MG cells using lipofectamine (GIBCO–BRL) according to the manufacturer's protocol. Firefly luciferase activity was assayed using a luciferase assay kit (Promega) according to the manufacturer's protocol. CAT activity was determined (Ausabel et al., 1995) and used to normalize the transfection efficiency. Promoter activity is defined as firefly luciferase activity (luminescence units) over CAT activity (c.p.m.). The activity exhibited by the fpten promoter in the sense orientation in pGL3-Basic was assigned as one arbitrary unit and the activities of the other plasmids are given relative to this.

Generation and analysis of stable fpten-expressing cell lines

U87-MG glioma cells (from ATCC) were transfected with pEGFP-C2-fpten or empty plasmid pEGFP-C2, and stable cell lines selected in medium containing G418 (250 μg/ml). The Fugu pten expression was examined by immunoblotting of cell lysates with anti-GFP antibody (Santa Cruz). For growth curve and saturation density determination, cells were cultured over a period of 2 weeks and the medium replaced every other day. Cells were counted at regular intervals using a haemocytometer. For the soft agar assay, cells were seeded in soft agar and maintained as described (Zheng et al., 1992). After 3 weeks, colonies were visualized with a phase contrast microscope and photographed.