Dolichol-Phosphate-Mannose-3 (DPM3)/Prostin-1 Is a Novel Phospholipase C-G Regulated Gene Negatively Associated with Prostate Tumor Invasion

Dolichol-phosphate-mannose-3 (DPM3)/prostin-1 is a novel phospholipase C-g regulated gene negatively associated with prostate tumor invasion

Elizabeth J Manos1, Marianne L-H Kim1, Jareer Kassis2, Pei-Yun Chang3, Alan Wells2 and David A Jones*,1

1Division of Molecular Pharmacology, Hunstman Cancer Institute, University of Utah, Salt Lake City, Utah, UT 84112, USA; 2Department of Pathology, University of Pittsburgh, and Pittsburgh VAMC, Pittsburgh, Pennsylvania, PA 15260, USA; 3Molecular and Cellular Pharmacology Program, University of Wisconsin-Madison, Madison, Wisconsin, WI 53706, USA

The most ominous development in tumor progression is death in men over age 50 (Mettlin 1993; Moul, 1996). the transition to an invasive and metastatic phenotype. Autopsy studies have revealed that, although localized Little is known, however, about the molecular alterations carcinomas exist in most elderly males, the majority of that cause a tumor to become invasive. In view of this, these were asymptomatic and medically unimportant we have used microarray expression analysis to evaluate (Whittemore et al., 1991). These localized lesions can be the expression pro®les of a unique panel of human cured by surgical or radiological interventions. Un- DU145 prostate cancer sublines that vary in their fortunately, once prostate cancer spreads beyond its invasive potential. The three DU145 sublines expressed capsular localization, current therapies only extend life epidermal growth factor (EGF) receptors that diered in by about a year, even in patients with minimal their ability to activate phospholipase C-g (PLCg). metastatic disease (Shultz, 1996). In order to design Three-way analyses yielded 11 genes out of 4608 genes rational therapies aimed at limiting prostate cancer screened that associated directly or inversely with progression we must ®rst de®ne key operative mole- invasive potential. The gene whose expression correlated cules in the various steps of this progression cascade. most strongly with lack of invasion was identi®ed as a Prostate cancer develops from epithelial cells that potential invasion suppressor and called prostin-1. escape the growth and dierentiation control of the Pharmacological inhibition of PLCg (U73122) con®rmed prostate microenvironment. These cells acquire proper- that PLCg signaling suppressed prostin-1 in that U73122 ties that enable them to penetrate surrounding base- treatment caused induction of prostin-1 in PLCg ment membranes, invade local tissue and grow competent cells. The prostin-1 gene, conserved through independently of the prostate environment as metas- phylogeny, is induced by androgen in LNCaP cells and tases. Tumor invasion is a distinct property of tumor encodes a 92 amino acid protein. The protein shares no progression as is evident by the ability of a variety of extensive homologies with other known genes, yet was agents to block invasiveness in vitro. For example, the recently identi®ed as a small stabilizer subunit of the tumor suppressor KAI blocks motility and invasion dolichol-phosphate-mannose (DPM) synthase complex. but not growth of colon cancer cells (Takaoka et al., That DPM3/prostin-1 might suppress tumor progression 1998), and the calcium channel inhibitor CAI prevents was supported by the ®nding that exogenous expression transmigration of Matrigel even in tumor cell lines in in COS cells leads to apoptosis. These ®ndings support which proliferation is unaected (Jacobs et al., 1997). the use of model cell lines to identify putative tumor Interestingly, estrogen can increase tumor growth even suppressors and promoters. Oncogene (2001) 20, 2781 ± while suppressing invasiveness (Rochefort et al., 1998). 2790. On the other hand, transforming growth factor-b (TGFb), which usually is anti-proliferative, can en- Keywords: prostate; invasion; metastasis; phospholi- hance tumor progression, possibly via its motility pase; EGF receptor; microarray promoting eects (Barrack, 1997). Finally, the distinc- tion between invasion and growth has been clearly demonstrated by an in vivo xenograft model of human Introduction prostate carcinoma in which disruption of PLCg signaling prevents tumor invasiveness but not tumor Cancer of the prostate is the most frequent malignant growth (Turner et al., 1996, 1997). tumor in men and is a leading cause of cancer-related Overexpression of EGF receptors in DU145 human prostate carcinoma cells promotes EGF receptor- dependent invasiveness both in vitro (Xie et al., 1995) and in vivo (Turner et al., 1996). Chen et al. (1994b) *Correspondence: DA Jones, Huntsman Cancer Institute, 2000 have demonstrated that EGF receptor-mediated moti- Circle of Hope, Rm. 5242, Salt Lake City, UT 84112, USA Received 1 November 2000; revised 7 February 2001; accepted 12 lity requires PLCg signaling by using a signaling- February 2001 restricted EGF receptor construct or pharmacological Identification of prostin-1 EJ Manos et al 2782 agents to block lipase activity of PLCg. In the absence carcinoma cell sublines that vary in their invasive of PLCg activity, haptokinetic motility is unaected potential because of dierential competence for PLCg (Chen et al., 1994b; Ware et al., 1998), and mitogenesis signaling. We approached this using high-density is actually increased by the lack of a competing cell cDNA microarrays to analyse the transcriptional response (Chen et al., 1996). Similar results have shown alterations coincident with onset of the invasive that PLCg is required for motility but not mitogenic phenotype. Our arrays displayed 4608 randomly signaling from receptors for platelet-derived growth selected, minimally redundant cDNAs in duplicate factor (PDGF) (Bornfeldt et al., 1995; Kundra et al., (Heller et al., 1997; Karpf et al., 1999; Schena et al., 1994a) and insulin-like growth factor (IGF-1) (Wenn- 1995, 1996). Figure 1a shows a schematic diagram of strom et al., 1994). Lastly, targeting PLCg signaling in the cell line comparisons. In experiment A, we mouse xenografts, utilizing either pharmacological compared the transcriptional pro®les of parental (U73122) or molecular (dominant-negative PLCz DU145 cells with DU145 cells stably overexpressing fragment of PLCg-1) agents, abrogates EGF mediated full length epidermal growth factor receptor (c'1186 invasiveness (Turner et al., 1996, 1997). Thus, PLCg EGFR). In experiment B we compared parental and subsequent downstream signaling events may DU145 cells with DU145 cells stably overexpressing a represent targets for dysregulated cell motility as truncated epidermal growth factor receptor incapable opposed to mitogenesis. of activating PLCg (c'973 EGFR) (Xie et al., 1995). One area drawing increasing attention is that of These sublines vary in their invasive capacity where altered transcription pro®les in motility and invasion. c'1186 EGFR4parental DU1454c'973 EGFR (Turner Whether these are due to genetic alterations or et al., 1996, 1997). Figure 1b shows a portion of an epigenetic responses to extracellular signals requires array from experiment A, parental DU145 versus further investigation. There appears to be a 4 ± 8 h c'1186 EGFR, and is typical of the array quality that induction period for EGF-induced motility that may we obtained. A detailed evaluation of the arrays was re¯ect the need for altered transcription (Maheshwari achieved by plotting the ratio of the hybridization et al., 1999). This requirement for de novo transcription signals (red signal/green signal) for each of the 4608 has been further implicated by the sensitivity of genes in the comparison. This analysis revealed that motility to transcription inhibitors (Bauer et al., 1992; most genes failed to respond to introduction of the full Chen et al., 1994a; Gordon and Staley, 1990). length EGFR and de®ned a control versus experi- Interestingly, urokinase-type plasminogen activator mental ratio of 1.0+0.172 (mean+s.d.) For the entire binding to its receptor and signaling via the ERK gene set. Expression ratios for individual points lying MAP kinases for only 30 min can stimulate mammary outside the standard deviation for the entire gene set carcinoma cell motility as measured over 24 h, identi®ed candidate, dierentially regulated genes. A suggesting a persistent change in the cell's transcription similar analysis was performed for experiment B, program (Nguyen et al., 1998). A few reports have parental DU145 versus c'973 EGFR expressing cells. pointed to the fos family of transcription factors as the Ratios from each comparison were then ranked from mediator of growth factor promoted motility and the most positive change to the most negative change invasion (Kustikova et al., 1998; Malliri et al., 1998). and those values lying at least two standard deviations In sum, there are solid indications that cell motility from the mean expression ratio were categorized requires altered transcription of select mRNA species according to their regulation in c'1186 EGFR expres- and although a few invasion genes have been described sing cells versus their regulation in c'973 EGFR using a candidate gene approach, the roster of speci®c expressing cells. elements involved in invasion remains incomplete. The changes we saw fell into three classes: (i) To expand our understanding of the events control- changes that were parallel in c'1186 EGFR and c'973 ling EGF mediated motility, and thus of PLCg signaling, EGFR expressing cells; (ii) changes that were unique to we have employed cDNA microarrays to compare c'1186 EGFR or c'973 EGFR expressing cells; and (iii) the transcriptional pro®les of 4608 genes in DU145 cells changes that were opposite in c'1186 EGFR and c'973 that express EGF receptors with dierential competency EGFR expressing cells. The majority of changes (8.5% for activation of PLCg. Our data illustrate the power of total) were shared between cells expressing c'1186 of syngeneic cells combined with high-throughput EGFR and cells expressing c'973 EGFR. This genomics in identifying candidate prostate invasiveness indicated that only a minority of the genes regulated genes. A novel PLCg-regulated gene, DPM3/prostin-1, by the EGF receptor requires an intact PLCg pathway. is presented as a new invasion suppressor. This simpli®ed the interpretation of our array data and demonstrated the strength of this model system in identifying those genes directly involved in the PLCg pathway and, therefore, in the control of EGFR Results mediated invasion. The second and third categories of changes contained a much smaller set of genes that Identification of putative invasion promoter and represent candidate invasion regulators. We were most suppressor genes interested in the third class of genes, those that showed We have examined the expression pro®les of 4608 opposite regulation in experiment A versus experiment randomly selected genes in three DU145 prostate B. We found only 11 genes (0.23% of total) that

Oncogene Identification of prostin-1 EJ Manos et al 2783

Figure 1 Hierarchical three-way analysis of candidate invasion genes using microarrays. (a) Messenger RNAs were harvested from each cell line and 1mg of each mRNA was reverse transcribed in the presence of Cy-3dCTP (parental) or Cy-5dCTP (c'1186 or c'973). Parental DU145 and either c'1186 (expt A) or c'973 EGFR (expt B) cDNA pools were combined and hybridized simultaneously against a microarray containing 4608 genes in duplicate. (b) Left panel. Shown is a single quadrant within the array from experiment A. Red hybridization signals indicate genes that are upregulated in c'1186 EGFR cells and green signals indicate genes that are down-regulated in c'1186 EGFR cells compared to parental DU145 cells. Yellow hybridization signals indicate equivalent expression levels in both cell lines. Right panel. Hybridization signal intensities from experiment A were captured using ArrayVision. Red and green signal intensities were normalized based on the total ¯uorescence of all the points in the red channel. Ratios of the red/green signals for each array element were calculated, averaged between the duplicate points and plotted against the array element number (each individual gene). As expected, most genes remain unchanged with a mean ratio of 1.0+0.172 (mean+s.d.). Points lying at least two standard deviations above or below the mean expression ratio were selected for comparison to those from experiment B conform to the criterion of having opposite expression that it showed the most complete suppression in c'1186 in c'1186 EGFR versus c'973 EGFR expressing cells as EGFR expressing cells and the most robust induction compared to the parental DU145 cells (Table 1). These in c'973 EGFR expressing cells. This gene fell into the were classi®ed as putative invasion promoters, if they candidate invasion suppressor category based on our were highest in c'1186 EGFR cells, or suppressors, if model system criteria. To con®rm the microarray they were highest in c'973 EGFR cells. correlation, we examined the level of this speci®c transcript in parental DU145 cells, c'1186 EGFR and c'973 EGFR expressing cells, as well as in DU145 cells Prostin-1 is highly expressed in non-invasive c'973 EGFR treated with the PLCg inhibitor U73122 (Figure 2). We and lacking in c'1186 EGFR cells found this gene to be most highly expressed in c'973 A single gene caught our attention during the EGFR expressing cells and U73122-treated cells. Its preliminary analysis (which we termed prostin-1) in expression was intermediate in the parental DU145

Oncogene Identification of prostin-1 EJ Manos et al 2784 Table 1 Genes identi®ed by microarray analysis that parallel the Prostin-1 corresponded to EST sequence AA047755 invasive phenotypes of DU145 cells and was a member of the unigene cluster Hs.110477. PLCg regulated genes Candidate promotera or suppressorb This unigene cluster contained 41 additional EST Urokinase receptor Promoter sequences that were overlapping and assembled into Insulin-like growth factor 1 Promoter two distinct contiguous sequences. The ®rst contiguous Hepatocyte growth factor Promoter sequence resulted in a clone of 525 base pairs that Platelet-derived growth factor-A Promoter contained a short open reading frame predicting a Cyclin I Promoter protein of 122 amino acids (molecular weight EST AA207900 Promoter EST548375 Promoter 13 275 Da). The second contig resulted in a clone of 70 kd heat shock protein Suppressor 374 bp that also contained an open reading frame. In MHC class 1 Suppressor this case, the predicted protein contained 92 amino EST AA548375 Suppressor acids with a molecular weight of 10 092 Da. The two EST AA047755 (DPM3/prostin-1) Suppressor sequences appear to dier through splicing of a 5 apromoter=c'11864Parental4c'973; bsuppressor=c'11865Parental prime exon that could result in a long and a short form 5c'973 of the protein (Figure 3). The shorter open reading frame is entirely contained within the longer contig in that the sequences of each contig were indentical from base pair position 19 in the short form through the poly(A)+ tail. Each contig contained a consensus polyadenylation signal. Based on the mRNA size for prostin-1 as seen in our Northern analysis (Figure 2), we concluded that we have obtained the full-length sequence for prostin-1 electronically. We obtained several of the clones (GenomeSystems, Inc.) and sequenced them to con®rm our informatics-based analysis.

Figure 2 DPM3/prostin-1 is regulated by the PLCg pathway. A Northern analysis of DPM3/prostin-1 was performed using mRNAs isolated from c'1186 EGFR expressing cells transfected with dominant negative PLCg (lane 1), DU145 cells treated with the PLCg inhibitor U73122 (lane 2), c'1186 EGFR expressing cells (lane 3), parental DU145 cells (lane 4), and c'973 EGFR expressing cells (lane 5). RNA levels were normalized before loading and equal amounts con®rmed by visualization of the 28S and 18S ribosomal subunits (lower panel). The DPM3/prostin-1 probe hybridized with a 490 bp mRNA based on comparison to RNA size standards. Signal intensities quanti®ed by phosphorimager analysis are indicated below as fold increase relative to c'1186 EGFR expressing cells

cells and suppressed in c'1186 EGFR expressing cells. In addition, introduction of a dominant-negative Figure 3 DNA sequence of human DPM3/prostin-1 and fragment of PLCg causes complete abrogation of deduced protein sequence. Assembly of the unigene cluster invasiveness in c'1186 EGFR expressing cells (Turner Hs.110477 resulted in nucleotide sequences of (a) 525 and (b) et al., 1997). Prostin-1 was highly expressed in these 374 bp. The sequences were con®rmed by direct sequencing of the cells. These results con®rmed the microarray data and clones corresponding to each contig. The sequences predicted proteins of 122 and 92 amino acids respectively. Genbank demonstrated that prostin-1 is regulated by PLCg accession numbers for the long and short forms of prostin-1 are signaling. AF312922 and AF312923 respectively

Oncogene Identification of prostin-1 EJ Manos et al 2785 Prostin-1 is a novel gene product 1996; Winberg et al., 1998). This same domain was also shared with cytoskeletal keratin, non-muscle myosin The putative proteins encoded by prostin-1 were novel heavy chain and paramyosin, all known cytoskeletal sequences and, therefore, required additional analyses proteins (Figure 4b). Given the importance of to form a hypothesis about their function. Using cytoskeletal rearrangement for cell motility (Lauen- BLAST (Altschul et al., 1990, 1997) and other protein burger and Horwitz, 1996; Stossel, 1993; Wells et al., motif searches, we found homologs of prostin-1 in 1998) and thus in the development of invasiveness and mouse, rat, C. elegans, Drosophila and yeast databases, metastasis, our identi®cation of prostin-1 as a novel indicating that prostin-1 is an evolutionarily conserved candidate invasion suppressor illustrates the potential protein (Figure 4a). The highest degree of conservation of microarray analyses to de®ne new molecular within this new protein family occurs within the C- determinants that accompany tumor progression to terminal region. The N-terminal region is predicted to invasiveness. While this manuscript was in preparation, contain a membrane-spanning region while the C- Maeda et al. (2000) disclosed the sequence of DPM3, a terminus is predicted to reside in the cytoplasm. We subunit of dolichol-phosphate-mannose synthase. The were unable to ®nd sequences other than human that short form of prostin-1 is identical to the sequence of encoded the long form of prostin-1 and, therefore, DPM3. While this initial identi®cation provides for concluded that the short form is more typical. possible functional roles, the exact role of DPM3/ Homology searches of prostin-1 against known prostin-1 in the synthase complex appears to be as an proteins revealed an interesting relationship to several adapter or stabilizer, and not one of enzymatic cytoskeletal proteins, typi®ed by plexins. The 27 function (Maeda et al., 2000). Thus, it remains to be residues at the C-terminus of prostin-1 showed determined whether this short peptide might also substantial homology (80%) with a domain of function in other cellular machineries. undescribed function contained in murine plexins, membrane proteins involved in neuronal cell contact, DPM3/prostin-1 is expressed ubiquitously and is axon guidance and fasciculation (Kameyama et al., androgen responsive We next examined the distribution of DPM3/prostin-1 within human tissues. Northern analyses revealed the expression of DPM3/prostin-1 in nearly all of the 16 normal tissues examined (Figure 5a). Similarly, seven of eight tumor cell lines examined demonstrated DPM3/prostin-1 gene expression (Figure 5b). Only Molt-4 cells lacked expression of DPM3/prostin-1. In order to ascertain the situation for dierential DPM3/prostin-1 expression, we next performed a virtual Northern using the Serial Analysis of Gene Expression Tag to Gene Mapping database curated by the NCBI (Lal et al., 1999; Velculescu et al., 1995; Zhang et al., 1997). At the time of analysis, the database oered 2 902 360 SAGE tags derived from 69 libraries representing diverse human tissue types. SAGE tags for DPM3/prostin-1 were found in 49 of the 69 libraries available. The highest number of DPM3/prostin-1 tags (345 tags/million) was seen in the ovarian cancer cell library, OVCA432-2. Next to this library, DPM3/prostin-1 tags were most abundant in libraries derived from the prostate cancer cell line LNCaP after treatment with androgen analogues (Figure 6a). In one case, tags corresponding to DPM3/prostin-1 increased nearly 10-fold after andro- Figure 4 Alignment of DPM3/prostin-1. (a) The predicted open gen treatment with R1881. In a second case, higher reading frame of human DPM3/prostin-1 was aligned with levels of DPM3/prostin-1 tags were observed in putative homologs from mouse, rat, C. elegans and yeast using LNCaP cells treated with dihydrotestosterone (DHT). the Clustal method in DNASTAR. Identical residues are shaded. The SAGE database contained only one set of libraries Conservation among the species is highest at the C-terminus with the N-terminus showing conserved hydrophobic, putative mem- derived from de novo normal prostate and prostate brane spanning regions. (b) A BLAST search against all known cancer. In this set, the prostate tumors were non- proteins revealed that the 27 amino acids at the C-terminus of invasive and showed higher numbers of DPM3/prostin- DPM3/prostin-1 share close homology with several cytoskeletal 1 tags than that seen in normal tissue (59 tags/million proteins. Shown in the alignment are murine plexins 1 and 2, human cytoskeletal keratin and human nonmuscle myosin heavy in the normal versus 130 tags/million in the tumor). chain (NMMHC). Identical or conserved residues between We then performed Northern analysis on RNA from DPM3/prostin-1 and aligned proteins are in bold LNCaP cells treated with DHT (Figure 6b). A

Oncogene Identification of prostin-1 EJ Manos et al 2786

Figure 5 Tissue distribution of DPM3/prostin-1. Expression of Figure 6 DPM3/prostin-1 is androgen inducible in LNCaP cells. DPM3/prostin-1 in (a) 16 normal human tissues and (b) eight (a) The distribution of tags corresponding to DPM3/prostin-1 was human cancer cell lines by Northern analysis. Hybridization examined in 69 SAGE libraries at the following web site: http:// signals were quanti®ed using a phosphorimager, normalized to www.ncbi.nlm.nih.gov/SAGE/SAGEcid.cgi?cid=110477) using GAPDH signals and are displayed in the plot as relative the Unigene number Hs. 110477. A plot of tags corresponding phosphorescence units to DPM3/prostin-1 in selected libraries is presented. These are compared to the average number of tags found in each of the 69 libraries available. (b) LNCaP cells were treated with vehicle of 10 nM DHT for indicated times. Poly(A)+ RNA was isolated and analysed by Northern blot using PCR-ampli®ed probes. Hybri- signi®cant increase in DPM3/prostin-1 expression was dization signals were quanti®ed using a phosphorimager, normal- seen at 12 and 24 h after treatment with DHT, ized to GAPDH, and displayed as relative phosphorescence units verifying the results from the SAGE analysis.

High level expression of DPM3/prostin-1 leads to compared to 4.9% (6/122) of cells expressing bovine apoptosis alkaline phosphatase as a control. Cells expressing Our analysis of DPM3/prostin-1 suggested it as a DPM3/prostin-1 underwent apoptosis and detached transmembrane protein with some homology to known from the culture plate, obscuring clear assessment of cytoskeletal proteins. To begin an analysis of the subcellular distribution. To overcome this, we exam- functional role of DPM3/prostin-1 in invasion we next ined the distribution of DPM3/prostin-1 in cells prior examined the subcellular distribution of DPM3/ to the onset of apoptosis. Figure 7a demonstrates the prostin-1 when overexpressed in Cos-1 cells. DPM3/ morphology and subcellular distribution of DPM3/ prostin-1 was tagged with a FLAG epitope and prostin-1 in Cos-1 cells at 24 and 48 h post-transfec- transfected into Cos-1 cells. Following 48 h of tion. DPM3/prostin-1 expressing cells showed signs of incubation cells were ®xed and stained with anti- apoptosis but at early time points revealed intense FLAG speci®c antibodies. Examination of cells expres- perinuclear localization of DPM3/prostin-1. This sing DPM3/prostin-1 at 48 h after transfection revealed staining pattern is consistent with localization of a high rate of cell death. Based on visual inspection, DPM3/prostin-1 in the endoplasmic reticulum and 71.5% (98/137) of cells expressing DPM3/prostin-1 con®rms observations presented by Maeda et al. displayed the apoptotic feature of condensed nuclei as (2000). Closer inspection revealed ®lamentous staining

Oncogene Identification of prostin-1 EJ Manos et al 2787 for FLAG-tagged DPM3/prostin-1 at sites distinct EGFR-stimulating loop is present in prostate carcino- from the perinuclear staining. This raises the possibility mas (Kim et al., 1999). The two EGFR constructs that DPM3/prostin-1 functions at other locations. (c'1186 and c'973) both activate the ERK MAP kinase (Chen et al., 1994b) and STAT pathways (David et al., 1996) but dier in that the c'1186 EGFR but not c'973 Discussion EGFR triggers the PLCg pathway (Chen et al., 1994a). This dierential restriction of a single de®ned pathway The power of any analysis is dictated by the model was instrumental in narrowing the focus and possible system investigated. Our previous work has determined candidates from pleuripotential signalers like EGFR. that growth factor-induced cell motility promotes One daunting problem with microarray analyses is prostate cancer progression. We have identi®ed PLCg that the large amount of data that one might generate signaling as a critical intracellular pathway that requires substantial downstream investigation. Our promotes invasion (Glading et al., 2000). These data show that array analyses of syngeneic cell lines ®ndings were obtained using a panel of sublines of that isolate a particular signaling intermediary allow the human prostate carcinoma-derived line DU145 that ®ltering of genes unrelated to the PLCg pathway and have dierent invasive potentials. These sublines were the invasive nature of these cells. If we just examine generated by expressing exogenously-encoded EGFR in genes that were found to be altered in a two-way cells that already presented a TGFa-EGFR autocrine comparison of either c'1186 EGFR or c'973 EGFR stimulatory loop (Xie et al., 1995). Such an autocrine expressing cells against parental DU145, we would have identi®ed approximately 8% of the genes as being altered by EGFR signaling. However, by presenting a hierarchical three-way comparison and limiting the signaling variable to a downstream event, PLCg activation, we were able to produce a manageable set of putative genes (*0.2% of examined genes). Further supporting the validity of this approach, genes surviving our analyses and ®lters included several with known promoter roles in tumor invasion. These included hepatocyte growth factor, insulin-like growth factor-1 (Boros and Miller, 1995), platelet-derived growth factor (Kundra et al., 1994a,b) and the urokinase receptor (Andreasen et al., 1997; Reuning et al., 1998; Schmitt et al., 1995). Each of these molecules has been implicated previously in tumor cell invasion and metastasis. The regulation of these genes by PLCg, however, is a novel observation. Although our array analysis revealed a number of genes with interesting potential functions in invasion, we focused on the characterization of a novel gene that showed the most convincing negative association with invasiveness. DPM3/prostin-1 appears to be encoded by two messenger RNAs that are generated by dierential exon splicing. These splice variants would produce two distinct products that dier with respect to the length of their N-terminus. Based on two pieces of evidence, we believe the shorter form of DPM3/ prostin-1 is expressed in most tissues. First, our Northern analyses failed to detect two message sizes Figure 7 Overexpression of DPM3/prostin-1 induces apoptosis in the tissues we have examined. The message size we in Cos-1 cells. (a) Cos-1 cells were transiently transfected with have detected is consistent with the shorter transcript. (left panels) FLAG-tagged DPM3/prostin-1 or (right panels) FLAG-tagged bovine alkaline phosphatase. At 24 h (upper Second, our search for DPM3/prostin-1 homologs panels) and 48 h (lower panels) following transfection, cells were demonstrated that rat, mouse, C. elegans and S. pombe ®xed and stained with anti-FLAG antibody. Primary antibody genomes all contain a DPM3/prostin-1 gene. In no case was detected using a FITC-conjugated secondary antibody and other than in human did we ®nd a homolog that visualized using ¯uorescence microscopy. At both 24 and 48 h post-transfection, BAP containing cells showed cytoplasmic corresponded to the long form of DPM3/prostin-1. staining and normal morphology. In contrast, DPM3/prostin-1 Nevertheless, the long form of the DPM3/prostin-1 expressing cells showed signs of detachment at 24 h and message is present in a number of clones from the condensed nuclei at 48 h. (b) Cos-1 cells transfected with dbEST database indicating that it is expressed. Further FLAG-tagged DPM3/prostin-1 showed a ®lamentous staining distinct from cytoplasmic BAP seen in (a). Superimposition of studies will be required to determine under what DPM3/prostin-1 staining with the nuclear stain DAPI is shown in conditions and in which tumor types the long form the right panel of the DPM3/prostin-1 transcript is produced and

Oncogene Identification of prostin-1 EJ Manos et al 2788 whether the translation start site produces the pleiotropic eectors, such as the EGF receptor and corresponding protein. similar receptors. Our identi®cation of DPM3/prostin-1 We found that the 92 amino acid peptide predicted as a novel, candidate invasion suppressor illustrates the for DPM3/prostin-1 was of novel sequence with some potential of microarray analyses in de®ning new C-terminal homology to certain cytoskeletal proteins, molecular determinants that accompany tumor pro- typi®ed by plexins. This small size of DPM3/prostin-1 gression to invasiveness. Continued application of suggested a role as a regulator rather than an enzyme. microarray expression technology to this problem This was con®rmed by Maeda et al. (2000) in that they oers the promise of expanding our list of candidate recently puri®ed DPM3/prostin-1 as a component of invasiveness genes and, therefore, our list of therapeu- the dolichol-phosphate-mannose synthase complex. tic targets. They demonstrated that DPM3/prostin-1 lacked direct enzymatic activity but rather stabilized the activity of DPM1 and DPM2 (Maeda et al., 2000). Interestingly, the domain in DPM3/prostin-1 that we found related Materials and methods to plexins and other cytoskeletal proteins was required for binding of DPM3 to the DPM synthase complex Cell culture (Maeda et al., 2000). DPM3 localized to the DU145 cells were propagated by standard methods (DMEM/ endoplasmic reticulum in human cells (Maeda et al., F12 media supplemented with 7.5% FCS) in the presence of 2000). This localization was con®rmed by our studies selection agents (1 mg/ml G418 for c'1186 and c'973 DU145 with DPM3/prostin-1 in Cos-1 cells, however, we plus 1.2 mg/ml methotrexate for PLCz-DU145 cells) (Chen et observed expression of DPM3/prostin-1 in sites distinct al., 1996; Turner et al., 1997). Cells were passaged in media/ FBS in the absence of selection agents for 3 days prior to from the ER. It is, therefore, possible that DPM3/ testing. LNCaP cells were grown in RPMI-1640 supplemen- prostin-1 also functions in other protein complexes. ted with 10% fetal bovine serum, 1.0 mM sodium pyruvate, We found that DPM3/prostin-1 expression nega- 2.0 mML-glutamine, 50 units/ml penicillin G sodium and tively associated with DU145 cell invasiveness, that 50 mg/ml streptomycin sulfate. Cells were grown at 378C PLCg signaling controlled its expression, and that under 5% CO2. Media and supplements were obtained from androgen could induce its expression in non-invasive Life Technologies, Inc. cell types. Each of these observations is consistent with a role of DPM3/prostin-1 and, perhaps the DPM Microarray experiments synthase complex, in maintaining a non-invasive state. One hypothesis emerging from the studies of Maeda et Microarray slides were prepared as previously described al. (2000) suggests a role for the DPM synthase in (Heller et al., 1997; Karpf et al., 1999; Schena et al., 1995, establishing levels of GPI-linked cell surface receptors. 1996). Total RNA was prepared from each experimental condition using Trizol reagent (Life Technologies, Inc.) This oers an intriguing possibility for the role of according to the manufacturer's instructions. Poly(A)+ DPM3/prostin-1 in invasiveness. Under normal phy- RNA was puri®ed from total RNA using the Oligotex-dT siological conditions, it is possible that DPM3/prostin- mRNA batch kit from Qiagen. Labeling and hybridization 1 helps maintain the levels of GPI-linked cell surface were performed as previously described (Heller et al., 1997; proteins, thus contributing to the maintenance of cell Karpf et al., 1999; Schena et al., 1995, 1996). Brie¯y, ®rst- surface interactions with the extracellular matrix. In strand cDNA probes were generated by incorporation of addition, overexpression of DPM3/prostin-1 caused Cy3-dCTP or Cy5-dCTP (Amersham Pharmacia) during detachment and apoptosis in Cos-1 cells while high reverse transcription of 1 mg puri®ed mRNA. The resulting levels of a control protein did not. This result may be cDNA probes were puri®ed by vacuum ®ltration, denatured due to disruption of cell surface interactions due to the at 948C and hybridized to an arrayed slide overnight at 428C. Slides were washed in 16SSC/0.2% SDS for 10 min, then in high levels of DPM3/prostin-1 protein. Further studies 0.16SSC/0.2% SDS for 20 min. Slides were rinsed, dried and are required to determine whether DPM3/prostin-1 has ¯uorescence was captured using the Avalanche dual laser a true pro-apoptotic role in vivo as well as to pursue confocal scanner (Molecular Dynamics). Fluorescent inten- the functional roles of DPM3/prostin-1 in invasion. sities were quanti®ed using Arrayvision 4.0 (Imaging Our studies, however, demonstrate that an unbiased Research). genomic approach, like microarray analysis, oers new hypotheses regarding the involvement of novel signal- Northern analyses ing pathways in biology and disease. Dysregulated signaling through PLCg does not Multiple human tissue mRNA blots purchased from promote the entire transition from a normal cell to Clontech were prehybridized at 658C for 1 h in Rapid-Hyb metastatic carcinoma. Rather this is likely but one key buer (Amersham Pharmacia), then hybridized with random- 32 signal in one step of tumor progression. However, it is primed, a- P-dCTP-labeled probe for 4 ± 16 h at 658C. Unbound probe was removed by washing twice in 26SSC/ necessary to narrow these types of open-ended searches 0.1% SDS at room temperature, then twice in 0.56SSC/ to enable one to study the resultant data points in 0.1% SDS at 658C. Blots were exposed to a phosphor screen depth, rather than selecting previously implicated genes and scanned using a Phosphorimager (Molecular Dynamics). from among a large pool of known and unknown LNCaP cells were plated at 106 cells/dish in 100 mm dishes candidates. This is especially likely to occur and be and allowed to adhere overnight. The next day, media was misleading when evaluating downstream signals from replaced with RPMI-1640 with 10% charcoal-stripped serum

Oncogene Identification of prostin-1 EJ Manos et al 2789 and grown for 24 h. Cells were then treated with vehicle or performed using FuGene6 (Roche) according to manufac- 10 nM dihydrotestosterone (Sigma) for various time points. turer's protocol. Brie¯y, cells were seeded into 4-well chamber Total RNA was isolated using Trizol (Life Technologies, slides (Nunc) and allowed to adhere overnight. FuGene6 was Inc.) and poly(A)+ mRNA was selected using Oligotex diluted into serum-free media (Opti-Mem, Life Technologies, (Qiagen). Two micrograms of poly(A)+ mRNA was run on Inc.) and incubated at room temperature for 5 min. Plasmid a denaturing formaldehyde agarose gel. RNA was transferred (200 ng) was then added to FuGene and allowed to incubate to nylon membrane (Amersham Pharmacia) and cross-linked. for 15 min. Transfection mix was added directly to cells and Blots were prehybridized at 428C for 1 h in UltraHyb buer incubated at 378Cin5%CO2 for 24 ± 48 h. Cells were (Ambion), then hybridized with random-primed, a-32P-dCTP- washed three times with TBS/Ca (50 mM Tris, pH 7.4, labeled probe for overnight at 428C. Unbound probe was 150 mM NaCl, 1 mM CaCl2), then ®xed with acetone/ removed by washing twice in 26SSC/0.1% SDS at 428C, methanol (1 : 1). After washing four more times with TBS/ then twice in 0.16SSC/0.1% SDS at 428C. Blots were Ca, slides were incubated with 10 mg/ml anti-FLAG M2 exposed to a phosphor screen and scanned using a monoclonal antibody for 1 h. Slides were washed ®ve times Phosphorimager (Molecular Dynamics). with TBS/Ca and then incubated with 10 mg/ml goat-anti- mouse secondary antibody conjugated to Alexa 488 (Mole- cular Probes, Inc.) for 1 h in the dark. Slides were washed Cell transfections and immunostaining four times with TBS/Ca, mounted with a coverslip using DPM3/prostin-1 was cloned into pFLAG-CMV-2 (Kodak) ProLong Antifade kit (Molecular Probes, Inc.), and observed for expression in Cos-1 cells. Transient transfections were using a ¯uorescence microscope.

References

Altschul SF, Gish W, Miller W, Myers WE and Lipman DJ. Kindra V, Soker S and Zetter BR. (1994b). Oncogene, 9, (1990). J. Mol. Biol., 215, 403 ± 410. 1429 ± 1435. Altschul SF, Madden TL, Schaer AA, Zhang J, Zhang Z, Kustikova O, Kramerov D, Grigorian M, Berezin V, Bock E, Miller W and Lipman DJ. (1997). Nucleic Acids Res., 25, Lukanidin E and Tulchinsky E. (1998). Mol. Cell. Biol., 3389 ± 3402. 18, 7095 ± 7105. Andreasen PA, Kjoller L, Christensen L and Duy MJ. LalA,LashAE,AltschulSF,VelculescuV,ZhangL, (1997). Int. J. Cancer, 72, 1 ± 22. McLendon RE, Marra MA, Prange C, Morin PJ, Polyak Barrack ER. (1997). Prostate, 31, 61 ± 70. K, Papadopoulos N, Vogelstein B, Kinzler KW, Straus- Bauer J, Margolis M, Schreiner C, Edgell CJ, Azizkhan J, berg RL and Riggins GJ. (1999). Cancer Res., 59, 5403 ± Lazarowski E and Juliano RL. (1992). J. Cell. Physiol., 5407. 153, 437 ± 449. Lauenburger DA and Horwitz AF. (1996). Cell, 84, 359 ± Bornfeldt KE, Raines EW, Graves LM, Skinner MP, Krebs 369. EG and Ross R. (1995). Ann. NY Acad. Sci., 766, 416 ± Maeda Y, Tanaka S, Hino J, Kangawa K and Kinoshita T. 430. (2000). EMBO J., 19, 2475 ± 2482. Boros P and Miller CM. (1995). Lancet, 345, 293 ± 295. Maheshwari G, Wells A, Grith LG and Lauenburger DA. Chen P, Gupta K and Wells A. (1994a). J. Cell. Biol., 124, (1999). Biophys. J., 76, 2814 ± 2823. 547 ± 555. Malliri A, Symons M, Hennigan RF, Hurlstone AF, Lamb ChenP,XieH,SekarMC,GuptaKandWellsA.(1994b).J. RF, Wheeler T and Ozanne BW. (1998). J. Cell. Biol., 143, Cell. Biol., 127, 847 ± 857. 1087 ± 1099. Chen P, Xie H and Wells A. (1996). Mol. Biol. Cell, 7, 871 ± Mettlin C, Jones GW and Murphy GP. (1993) CA Cancer J. 881. Clin., 43, 83 ± 91. David M, Wong L, Flavell R, Thompson SA, Wells A, Moul JW. (1996). Fed. Pract., 13, 3±8. Larner AC and Johnson GR. (1996). J. Biol. Chem., 271, Nguyen DH, Hussaini IM and Gonias SL. (1998). J. Biol. 9185 ± 9188. Chem., 273, 8502 ± 8507. Glading A, Chang P, Lauenburger DA and Wells A. (2000). Reuning U, Magdolen V, Wilhelm O, Fischer K, Lutz V, J. Biol. Chem., 275, 2390 ± 2398. Grae H and Schmitt M. (1998). Int. J. Oncol., 13, 893 ± Gordon SR and Staley CA. (1990). Cell. Motil. Cytoskeleton, 906. 16, 47 ± 57. RochefortH.PlatetN,HayashidoY,DerocqD,LucasA, Heller RA, Schena M, Chai A, Shalon D, Bedilion T, Cunat S and Garcia M. (1998). J. Steroid. Biochem. Mol. Gilmore J, Woolley DE and Davis RW. (1997). Proc. Natl. Biol., 65, 163 ± 168. Acad. Sci. USA, 94, 2150 ± 2155. Schena M, Shalon D, Davis RW and Brown PO. (1995). Jacobs W, Mikkelsen T, Smith R, Nelson K, Rosenblum ML Science, 270, 467 ± 470. and Kohn EC. (1997). J. Neurooncol., 32, 93 ± 101. Schena M, Shalon D, Heller R, Chai A, Brown PO and Davis Kameyama T, Murakami Y, Suto F, Kawakami A, Takagi S, RW. (1996). Proc. Natl. Acad. Sci. USA, 93, 10614 ± Hirata T and Fujisawa H. (1996). Biochem. Biophys. Res. 10619. Commun., 226, 396 ± 402. Schmitt M, Wilhelm O, Janicke F, Magdolen V, Reuning U, Karpf AR, Peterson PW, Rawlins JT, Dalley BK, Yang Q, OhiH,MoniwaN,KobayashiH,WeidleUandGraeH. Albertsen H and Jones DA. (1999). Proc. Natl. Acad. Sci. (1995). J. Obstet. Gynaecol., 21, 151 ± 165. USA, 96, 14007 ± 14012. Shultz MJ. (1996). Fed. Pract., 13, 14 ± 18. Kim HG, Kassis J, Souto JC, Turner T and Wells A. (1999). Stossel TP. (1993). Science, 260, 1086 ± 1094. Histol. Histopathol., 14, 1175 ± 1182. Takaoka A, Hinoda Y, Satoh S, Adachi Y, Itoh F, Adachi M Kundra V, Escobedo JA, Kazlauskas A, Kim HK, Rhee SG, and Imai K. (1998). Oncogene, 16, 1443 ± 1453. Williams LT and Zetter BR. (1994a). Nature, 367, 474 ± Turner T, Chen P, Goodly LJ and Wells A. (1996). Clin. Exp. 476. Metastasis, 14, 409 ± 418.

Oncogene Identification of prostin-1 EJ Manos et al 2790 Turner T, Epps-Fung MV, Kassis J and Wells A. (1997). Whittemore AS, Keller JB and Betensky R. (1991). J. Natl. Clin. Cancer Res., 3, 2275 ± 2282. Cancer Inst., 83, 1231 ± 1235. Velculescu VE, Zhang L, Vogelstein B and Kinzler KW. Winberg ML, Noordermeer JN, Tamagnone L, Comoglio (1995). Science, 270, 484 ± 487. PM, Spriggs MK, Tessier-Lavigne M and Goodman CS. Ware MF, Wells A and Lauenburger DA. (1998). J. Cell. (1998). Cell, 95, 903 ± 916. Sci., 111, 2423 ± 2432. Xie H, Turner T, Wang MH, Singh RK, Siegal GP and Wells Wells A, Gupta K, Chang P, Swindle S, Glading A and A. (1995). Clin. Exp. Metastasis, 13, 407 ± 419. Shiraha H. (1998). Microsc. Res. Tech., 43, 395 ± 411. Zhang L, Zhou W, Velculescu VE, Kern SE, Hruban RH, Wennstrom S, Siegbahn A, Yokote K, Arvidsson AK, Hamilton SR, Vogelstein B and Kinzler KW. (1997). Heldin CH, Mori A and Claesson-Welsh L. (1994). Science, 276, 1268 ± 1272. Oncogene, 9, 651 ± 660.

Oncogene