EGF- and Cell-Cycle–Regulated STAG1/PMEPA1/ERG1.2 Belongs to a Conserved Gene Familyand Is Overexpressed and Amplified inBreast and Ovarian Cancer

Giuseppe Giannini,1* Maria Irene Ambrosini,2 Lucia Di Marcotullio,1 Fabio Cerignoli,2 Massimo Zani,1

Andrew Ray MacKay,2 Isabella Screpanti,1,3 Luigi Frati,1,4 and Alberto Gulino1,4

1Department of Experimental Medicine and Pathology, University La Sapienza, Rome, Italy2Department of Experimental Medicine, University of L’Aquila, L’Aquila, Italy3Pasteur Institute, Cenci-Bolognetti Foundation, Rome, Italy4Neuromed Institute, Pozzilli, Italy

The abnormal activation of the epidermal growth factor (EGF) pathway is one of the most common findings inhuman cancer, and a number of molecular devices of laboratory and clinical relevance have been designed to blockthis transduction pathway. Because of the large number of cellular events that might be regulated through the

activation of the four EGF receptor family members, it is possible that screening methodologies for the identification ofnew molecular targets working downstream of these pathways may provide new tools for cancer diagnosis andpotentially prevention and therapy. In searching for EGF target genes, we have identified ERG1.2, the mouse homolog

of the solid tumor-associated gene STAG1. Both in humans and in mice, it belongs to a new gene family that can giveorigin to several protein isoforms through alternative splicing and/or multiple translation starts. Sequence analysis andexperimental data suggest that ERG1.2 is likely to function as a membrane-bound protein interacting withdownstream signaling molecules through WW- and SH3-binding domains. ERG1.2 is a cell-cycle–regulated gene, and

both ERG1.2 and STAG1 are induced by EGF and other growth factors at the transcript and protein levels. Finally, wehave demonstrated that, besides prostate cancer and renal cell carcinoma, STAG1 was also overexpressed in breastand ovarian cancer cell lines and in breast primary tumors. Although in most cases STAG1 overexpression is probably

due to the abnormal activation of the EGF pathway, we have also demonstrated genetic amplification andrearrangement of its locus in one breast cancer cell line and one primary ovarian cancer, suggesting that STAG1 mightbe a direct molecular target in the carcinogenetic process. Thus its overexpression might be regarded not only as a

tumor marker but also as a potentially pathogenetic event. � 2003 Wiley-Liss, Inc.

Key words: growth factor; cell cycle regulation; breast cancer; ovarian cancer

INTRODUCTION

Signaling through the epidermal growth factor(EGF) receptor family is involved in a large array ofcellular functions including proliferation, survival,adhesion, migration, and differentiation. In addi-tion to participating in organmorphogenesis, main-tenance, and repair, upregulated and deranged EGFRsignaling is observed in human cancer [1]. This isbest exemplified by the overexpression of the ErbB2receptor in breast and ovarian cancer. In particular,in breast tumors ErbB2 is overexpressed because ofgene amplification in 20–30% of cases and is consi-dered an important prognostic factor, predictive ofmore aggressive tumor behavior and poor outcome[2]. Also, overexpression and structural alteration ofErbB1 are rather frequent in human malignanciesand account for more than 40% of cases of humangliomas [3,4].Derangementof the signaling initiatedby the EGF receptor family may also require ordepend on the expression/overexpression of specific

ligands. For example, TGF-a expression correlateswith poor prognosis in patients with lung, ovary andcolon cancer and a paracrine loop involving neur-egulin-1 and ErbB2/ErbB3 receptors was described inprostate cancer [5,6]. Overall, the aberrant activationof signaling events downstream of the different EGFreceptor familymembers is probably one of themostcommon themes in human cancers. Thus, althoughmuch is known about the EGFR signaling events

MOLECULAR CARCINOGENESIS 38:188–200 (2003)

� 2003 WILEY-LISS, INC.

*Correspondence to: Department of Experimental Medicine andPathology, University La Sapienza, Policlinico Umberto I, VialeRegina Elena, 324, 00161, Rome, Italy.

Received 17 March 2003; Revised 26 August 2003; Accepted 13October 2003

Abbreviations: EGF, epidermal growth factor; STAG1, solid tumorassociated gene 1; ERG, EGF responsive gene; FBS, fetal bovineserum; CHX, cycloheximide.

DOI 10.1002/mc.10162

involved in cancer development and progression,screening methodologies for the identification ofnewmolecular targets working downstream of thesepathways are likely to provide new tools to be usedfor cancer diagnosis and potentially prevention andtherapy.Wehave previously established the TC-1S cell line,

a mouse epithelial cell culture derived from thymicstroma that may develop a neurotypic phenotype inresponse to EGF [7] and other growth factors [8].Recently, we have used this cell system to identifynew targets of EGF involved in mediating its mito-genic and differentiative effects, through a differen-tial display/RNA fingerprinting protocol [9,10]. Herewe report that ERG1.2, an EGF responsive gene, is themouse ortholog of the recently described cancer-associated STAG1/PMEPA1gene (henceforth referredto as STAG1). The new gene family it belongs tocomprises at least two genes coding for several iso-forms, both in humans and mice. Sequence analysissuggests that they probably function as membrane-boundproteins interactingwith downstream-signal-ing molecules. We have also shown that ERG1.2/STAG1 is a cell-cycle–regulated gene induced by EGFand other growth factors at the transcript and pro-tein level. Finally,wehave demonstrated that STAG1was overexpressed in breast and ovarian cancer celllines and in breast primary tumors. Furthermore, itwas amplified in a breast cancer cell line and its locuswas rearranged and amplified in a primary ovariancancer.

MATERIALS AND METHODS

RNA and Protein Expression Studies

Following treatment, cells were scraped from thedish, washed twice with ice-cold PBS, and processedfor RNA extraction with the RNAeasy kit (Quiagen,Inc., Chatsworth, CA). For Northern analysis, totalRNA (20 mg) were subjected to a gel electrophoresis,blotted to nylonmembranes (Gene Screen Plus, NENLife Science Product, Boston, MA), and hybridizedto randomly primed cDNA probes. Multiple tissueNorthern blots (Clontech, Palo Alto, CA) containedpolyadenilated RNA (2 mg) from various tissues. Forprotein extraction, cells were washed and lysed inRIPA buffer (1% NP-40; 0.5% sodium deoxycholate;0.1% SDS; 140 mM NaCl; 50 mM Tris-HCl, pH 7.6;5mMEDTA) supplementedwithprotease inhibitors.After that, cell lysates were passed through a 21-gauge needle and cleared by centrifugation at15000� g for 20 min at 48C. After normalizationwith BCA protein assay (Pierce, Rockford, IL), eachprotein lysate was separated on SDS–PAGE, blottedontonitrocellulosemembrane (Schleicher&Schuell,Dassel, Germany), and probed with the 2B#8 rabbitpolyclonal antibody diluted to 0.4 ng/mL or withthe anti-c-myc (sc-40) mouse monoclonal antibody(Santa Cruz Biotechnology, CA) diluted 1:1000. Im-

munoreractive bands were visualized by enhancedchemiluminescence (Pierce).

Production of Anti-ERG1.2 Polyclonal Antibody

Polyclonal antibodies against amino acid (aa)residues 143 to COOH-ter of recombinant GST-ERG1.2 fusion protein were raised in New Zealandwhite male rabbit and purified using Sephadex Fastflow beads conjugated with a Protein A (Sigma, UK)[11].

In Vitro Transcription and Translation

The plasmids containing the myc-tagged and theHA-tagged e fragment of ERG1.2 (m-e and HA-e) andthe myc-tagged STAG1b (m-MGV) were linearizedand used as templates in the TNT T7 coupledreticulocyte lysate system (Promega Corporation,Madison, WI) in the presence of 35S-methionine(Amersham Pharmacia Biotech, Piscataway, NJSJ1015), according to the supplier’s instructions.

Cell Culture and Transfection

The SK-OV-3, OVCAR-3, IGROV-1, INTOV-1, andTC-1S cell lines were cultured in RPMI 1640mediumsupplemented with 10% fetal bovine serum (FBS),glutamine and antibiotics, plus 0.1 mg/mL sodiumpyruvate (onlyTC-1S) and50mM2-mercaptoethanol(only INTOV-1). MCF10A cells were grown in a1:1 mixture of Ham’s F12 and DMEM supplement-ed with glutamine and antibiotics, 20 ng/mL EGF,500 ng/mL hydrocortisone, 0.01 mg/mL insulin, 5%FBS. The remaining cell lineswere cultured inDMEMsupplemented with 10% FBS, glutamine, and anti-biotics. Transient transfections were performed withTransfast reagent (Promega Corporation) accordingto the manufacturer’s instructions. For stimulationstudies, subconfluent cells were treated with EGF(10 ng/mL), IGF (30 ng/mL), PDGF (50 ng/mL), TPA(30 ng/mL) (all from Collaborative Res., Inc., Bed-ford, MA), or cycloheximide (10 mg/mL, Sigma, UK).Inhibition of EGF receptor pathway was performedon BT-20 cells with PD168393 (2 mM) or AG1478(5mM).Forproteasome inhibition studies,C3H10T1/2 cells were treated overnight with MG-132 (Calbio-chem-Novabiochem Corp., La Jolla, CA, 10 mM),whereas HaCat cells were treated with MG-132(3.3 mM ) for 8 h. For glycosylation studies, NIH3T3cells were transfected with an STAG1b expressingplasmid and treated with Tunicamycin or benzyl-a-GalNAc (6 mg/mL and 2 mM, respectively, Calbio-chem-Novabiochem Corp.) or both for 24 h.

Pull Down Assays

The E. coli BL21 strain was induced to express theGST-fusion constructs for 2 h with 0.5 mM IPTG.Cells were then harvested and resuspended in100 mM NaCl NETN buffer (20 mM Tris pH 8;100 mM NaCl; 0.5% NP-40; 1 mM EDTA) plusprotease inhibitors. After sonication, the extracts

STAG1/PMEPA1/ERG1.2 IN BREAST AND OVARIAN CANCER 189

wereclarifiedbycentrifugation.The recovered super-natants were allowed to bind to Glutathione Sephar-ose 4B beads (Amersham) for 1 h at 48C. Afterwashing in 700 mM NaCl NETN buffer, the beadswith the ligated GST-fusion constructs were resus-pended in the interaction buffer (50mMTris pH 7.4;150 mM NaCl; 0.1% NP-40; 2 mM EDTA) andincubated with 35S-STAG1b (1/5 of the translationreaction) for 2 h in rotation at 48C.After fivewashingsteps in the interaction buffer, the beads wereresuspended and boiled in 5� SDS sample buffer(20 ml) twice, and the supernatants were collected.An aliquot of each sample was loaded onto SDS–PAGE. The gels were then stained with Coomassieblue, treated with Amplify (Amersham), dried, andexposed.

DNA Extraction and Analysis

For Southern analysis, genomic DNA from eachcell line (30 mg) were digested with the appropriaterestriction enzymes, electrophoretically separated in0.8% agarose gel, and transferred to nylon mem-branes (GeneScreen Plus, NEN). Membranes werehybridized with a [32P]-labeled DNA probe, obtainedby EcorI/PvuII digestion of the AA027926 EST clone,homologous to STAG1.

RESULTS

Identification, Cloning, Sequence Assembly,and Genomic Organization of ERG1.2

RNA fingerprinting analysis of EGF-treated TC-1Scells resulted in the identificationof anumber of EGFresponsive genes (ERGs) [10], five of which did notrecognize functionally characterized homologs inthe databases. Northern hybridization of the differ-entially displayed 1.2/17 fragment (dd1.2/17) onTC-1S poly(A)þ RNA identified three distinct transcripts(Figure 1A), the 2.4 kb form being the mostabundant. EGF induced all of them in stimulatedcells (Figure 1A).A cDNA contig generated by two overlapping

phage inserts (ld and le) isolated from amouse lungcDNA library encompassed 1922 bp before thepoly(A) tail (Figure 1B) and recapitulated the com-plete nucleotide sequence of the 2.4 kb transcript.The nucleotide sequence (Figure 1D) revealed thepresenceof anopen readingpotentially starting fromtwo different in-frame ATG (Figure 1D) that wouldgive predicted protein sequences of 274 and 245 aa,respectively. Besides a typical AATAAA polyadenila-tion signal, the 30 untranslated region containedseveral ATTTA and reach sequences potentiallyinvolved in the destabilization of the transcripts[12] and theTTTTGTAmotifs previously shown to beinvolved in controlling the expression of a numberof immediate-early genes [13]. The organization ofERG1.2 locus schematically represented in Figure 1Cwas deduced by comparing our cDNAwith the geno-

mic sequences of a public supercontig (MGSGV3mouse genome database accession no. NW_000180)containing approximately 50 kb of the ERG1.2 locusand a genomic phage clone we isolated (not shown)and containing the exon 1. From this analysis, wealso deduced that ERG1.2 maps on the terminal partof mouse chromosome 2, close to Bmp7, Pck1, andGnas, in a region synthenic to the human chromo-some 20q13.3.

ERG1.2 Protein

ERG1.2 predicted protein sequence appearedhighly homologous (84% identity) to the predictedprotein products of the human STAG1, a cancer-relatedgene ([14,15] and seebelow), toC18orf1 (61%identity), the predicted protein product of a geneidentified on chromosome 18 [16], and to its mousehomolog (mC18orf1, 59% identity)wehave recentlycloned (unpublished observation) (Figure 2A), thuspotentially identifying a new protein family. Com-putational analysis of ERG1.2 protein sequencewith PIX (http://www.hgmp.mrc.ac.uk/Registered/Webapp/pix/) confirmed the presence of a potentialtype 1b transmembrane domain between aa 37 and58, a region highly conserved also in C18orf1 andSTAG1. It also suggested the presence of interestingmolecular features and domains in the differentfamily members as summarized in Figure 3A. Theyinclude potential SH3 and WW binding domains inthe proline reach carboxyterminal region and glyco-silation, myristoylation, and phosphorylation sitesfor protein kinase C and casein kinase-2 (Figure 2A).We also noticed two largely divergent regions in thecarboxyterminal tail of the proteins, suggesting thatthey may contain regulatory sites conferring speci-ficity on each molecule. In a third very divergentregion located in the N-terminus, C18orf1a con-tained a low-density lipoprotein receptor class Adomain (LDLRA) thatmay potentially confer ligand-binding capabilities on this protein. C18orf1b,which derives from the same gene through analternative splicing, lacks this domain [16]. Througha similar mechanism, STAG1 is predicted to giveorigin to two proteins [14], only one of which,STAG1b, contains anadditional regionN-terminal inits transmembrane domain (Figure 2A). In vitrotranslation of ERG1.2 confirmed the presence oftwo active translation starts (Figure 2B, m-MGV),being the most abundant protein compatible in sizewith a first ATG start. Although the shorter proteinmight have been generated through a TNT artifact, itmore probably originates from the second ATG thatis surrounded by a strongly conserved Kozac con-sensus (Figure 1D). Indeed, a cDNA construct lackingthe first ATG was very efficiently translated bothin vitro and in vivo (Figure 2B, m-e). Expressed inNIH3T3 cells, several constructs generated proteinswith reduced electrophoretic mobility compared tothe in vitro–translated ones (Figure 2B), indicating

190 GIANNINI ET AL.

that some of the predicted post-translational mod-ification/s occur in the living cell. In particular,transfection of the full-length ERG1.2 construct(Figure 2B, m-MGV) or STAG1b (Figure 2C) resultedin the appearance of a ladder of bands between 35and 40 kDa. In contrast, cells transfected with eithera shorter version of the ERG1.2 (Figure 2B, m-e) or

STAG1a (not shown) showed a distinct band ofapproximately 35 kDa in size. Interestingly, STAG1band ERG1.2, but not STAG1a or m-e, shared thepresence of two N-glycosylation sites before thetransmembrane region. Treatment of the transfectedcells with tunicamicin thatmostly inhibitsN-linked-glycosylation, strongly reduced the size of the

Figure 1. cDNA sequence and genomic organization of ERG1.2.(A) The differentially displayed dd1.2/17 fragment recognizes threedifferent transcripts of approximately 5, 2.4, and 1.9 kb on Poly(A)þRNA from EGF-treated TC-1S cells; (B) ERG1.2 full-length cDNA wasassembled by sequencing two distinct phage clones obtained by amouse lung cDNA library screening; (C) Structure of ERG1.2 gene onthe mouse chromosome 2; exon and intron sizes are indicated. (D)

cDNA sequence of ERG1.2. Kozac consensus sequences surroundingthe two possible starting methionine are indicated (gray boxes);translation is given below; intron/exon boundaries are marked witharrowheads, the polyadenilation signal in bold; the AT reachsequences are underlined; and the TTTTGTA motif in the 30-UTR isdouble underlined.

STAG1/PMEPA1/ERG1.2 IN BREAST AND OVARIAN CANCER 191

STAG1b-transfected protein (Figure 2C), thus sup-porting the hypothesis that those glycosylation sitesare used in vivo (Figure 2C). Finally, we tested thecapability of the carboxyterminal proline reachregion of ERG1.2 and STAG1b to mediate protein-protein interaction, by using SH3 orWWcontainingproteins in GST-pulldown experiments. YAP65 WWdomain 1 and 2, NEDD4 WW-domain 2, and GRB-2SH3 domain clearly interacted with the in vitrotanslated STAG1 andERG1.2 proteins (Figure 2Danddata not shown). Under the same conditions, STAG1failed to bind to spectrin, NCK-1, and SRC SH3domains and to FE65 and PIN1 WW domains.Further confirming the specificity of the interaction,STAG1 did not bind to amutant version of YAPWWdomain 1(Figure 2D, WW1-M) [17].Overall, these data indicate that ERG1.2 is the

mouse homolog of STAG1 and belongs to a family of

conserved membrane proteins characterized by aproline-reach region which can mediate protein-protein interaction; at least half of the members ofthe family might potentially interact with specificligands through either an LDLRA domain or anheavily glycosylatedN-terminal domain, to generatespecific signals.

Expression Pattern of ERG1.2 and STAG1

To gain information on the pattern of expressionofERG1.2,weanalyzedapanelofmouse tissues andanumber ofmouse cell lines byNorthern blotting.Wedetected ERG1.2 expression in all mouse tissues andinmost cell lines analyzed (Figures 3–6 and Table 1),the 2.4-kb being the most abundant species. Arelevant exception was represented by the absenceof any isoform in liver and testis and in the FLC cells(Figure 3A and Table 1). Interestingly, EGF induced

Figure 2. ERG1.2 protein family structure and functional sites. (A)Alignment of the different isoforms belonging to the ERG1.2 family.Predicted PKC phosphorylation (*.*), CK2 phosphorylation (*..*),myristoylation (MYR), and glycosilation (GLY) sites are indicated;TMD, conserved transmembrane domain type 1b; LDLRA, lowdensity lipoprotein receptor A domain; DR1, 2, 3, divergent regions1, 2, 3; SH3 binding domains are in gray boxes; WW binding domainsare in dashed boxes. (B) Several ERG1.2 constructs (m-e and H-e,myc-tagged and HA-tagged version of the e-fragment, indicated in

Figure 1B; m-MGV, myc-tagged ERG1.2 full-length cDNA) wereeither in vitro translated or transfected in NIH3T3 cell and theirtranslation was revealed by the anti-tag detection. (C) NIH3T3 cellswere transfected with a full-length STAG1b construct and sub-sequentially treated with tunicamicin (TUN) and/or benzyl-a-GAL-NAC (BEN) and processed for Western blotting. (D) The indicatedGST-fused WW- and SH3-domains expressed in E. Coli were used topull-down 35S-labeled in vitro translated STAG1b. Coomassiestaining of the GST-fused peptides is given below.

192 GIANNINI ET AL.

a large increase in the accumulation of ERG1.2transcripts in fibroblast and epithelial cells withlower constitutive expression, whereas it was lesseffective in cells with higher constitutive expression,as in the case of B104 and TS-A, a neural and a breastepithelial cancer cell line, respectively. We alsoinvestigated STAG1 expression on a panel of humantissues and cell lines. As previously reported [15], wefound the highest expression of STAG1 in prostate.However, we noticed a lower but widely diffusedexpression in most tissues (Figure 3B) includingbreast (Figure 7C) andovarian epithelium (Figure 3B)also. As for mouse ERG1.2, STAG1 was also basicallyundetectable in liver and peripheral blood leuko-cytes (Figure 3B). Most tissues expressed only the5- and 2.4-kb forms, whereas in prostate we alsodetected the 1.9-kb form.More clearly, STAG1 show-ed three distinct transcripts of 5, 2.4, and 1.9 kb inmany cultured cell lines (Table 2 and Figure 3C),although we did not notice an overt prevalence ofthe 2.4-kb form (Table 2, Figures 3 and7), as observedin the mouse cells. Similar to the murine ERG1.2,STAG1 was regulated by EGF in Ha-Cat and Saos-2cell lines (Figure 3C) and in a number of otherfibroblast and epithelial cell lines, includingmammary and ovarian epithelial cells (Figure 7 andTable 2).

Figure 3. Expression pattern of ERG1.2 and STAG1. Mouse (A)and human (B) multiple tissue and EGF-treated human cell lines (C)Northern blots were hybridized with either ERG1.2 (A) or STAG1 (B,C) probes. GAPDH hybridization or ethidium bromide staining of the

28S RNA were used as gel loading control. Black arrowheads indicatethe three ERG1.2 or STAG1 transcripts; white arrowhead specificallyindicates the 1.9 kb transcript in prostate.

Figure 4. ERG1.2 mRNA expression is growth factor and cell-cycleregulated in TC-1S cells. (A) Detailed time-course of ERG1.2expression in response to EGF in TC-1S cells. (B) Effect ofcycloheximide (CHX) on the accumulation of ERG1.2 mRNA in TC-1S cells. (C) Following growth arrest obtained by contact inhibitionand serum starvation in 0.5% FBS for 48 h, TC-1S cells were replatedin the presence of 10% FBS for the indicated time and total RNA wasextracted and analyzed by Northern blotting.

STAG1/PMEPA1/ERG1.2 IN BREAST AND OVARIAN CANCER 193

Induction of ERG1.2 mRNA Is AssociatedWith Growth Stimulation

In TC-1S cells, ERG1.2 mRNA accumulationstarted after 1 h, peaked between 3 and 6 h of EGFtreatment, and returned to basal levels after 24 h(Figure 4A). Cycloheximide (CHX) treatment didnot modify the EGF-dependent induction of ERG1.2after 1 h, and CHX on its own did not alter theconstitutive level (Figure 4B), thus indicating thatEGF directly induces ERG1.2 expression. At latertime-points, however, CHX induced accumulationof ERG1.2 transcript, either alone or in the presenceof EGF suggesting that continuous protein synthesisis required for the physiological processing ofERG1.2 mRNA (Figure 4B). We then arrested TC-1Scells bybothcontact inhibitionand serumstarvationfor 48 h. After this time, the vast majority of cellswere in the G1 phase and fewer than 2% in S phase(not shown), confirming that TC-1S cells entered aquiescent state. Alongwith becoming quiescent, TC-1S cells reduced the expression of ERG1.2, whichwasminimal in confluent cultures left in 0.5% FBS for48 h. Between 15 and 21 h after releasing TC-1S

cells from quiescence, most cells had completedS phase (not shown). In contrast, already after 1 h ofserum addition, we observed a strong increase inERG1.2mRNA steady state levels, whichpeaked after4 hbefore decreasing at later time-points (Figure 4C).These results indicate that ERG1.2 is cell-cycle–regulated in TC-1S cells. We then tested whetherERG1.2 is subjected to the same regulation infibroblasts also. We obtained similar results withCHX (Figure 5A) and we confirmed that a lowerERG1.2 expressionwas associatedwith the quiescentstate in NIH3T3 and Balb3T3 fibroblasts, and itsexpression increased as cells were stimulated withserum (Figure 5B and D). Other growth-promotingfactors, including IGF, PDGF, and the phorbol estherTPA also induced ERG1.2 expression in eitherasynchronous (Figure 5C) or quiescent (not shown)NIH3T3 cells. At variance with NIH3T3, serum-starved Balb3T3 cells failed to re-enter cell cyclewhen stimulatedwith either EGFor IGF.Under theseconditions, they also failed to induce ERG1.2 ex-pression (Figure 5D). In contrast, ERG1.2 expressiontransiently increased after PDGF treatment, which,even alone, was able to stimulate the cells to re-enter

Figure 5. ERG1.2 mRNA expression is growth factor–and cell-cycle–regulated in fibroblasts. (A) Effect of CHX on the accumulationof ERG1.2 mRNA in NIH3T3 cells. (B) Following serum starvation in0.5% FBS for 48 h, NIH3T3 cells were stimulated with 10% FBS forthe indicated time and total RNA was extracted and analyzed byNorthern blotting. (C) Exponentially growing NIH3T3 cells werestimulated for the indicated time with EGF, IGF, PDGF, and TPA, andtotal RNA was extracted and analyzed by Northern blotting. (D)Balb3T3 cells were serum-starved for 24 h and subsequently

restimulated with either 10% FBS or IGF, PDGF, and EGF for theindicated time before harvesting cells for cell-cycle FACS analysis orRNA preparation. G0/G1, S, and G2/M phase percent values are givenin the table. Only cells treated with FBS or PDGF entered the cell cycleas demonstrated by the increase in S and G2/M cells 18 and 40 h afterstimulation and by the expression of the S-phase gene MCM2. Weonly observed a transient increase in ERG1.2 expression in FBS-stimulated cells and to a lesser extent, in PDGF-stimulated cells.Vimentin hybridization was used as gel loading control.

194 GIANNINI ET AL.

the cell cycle, as shown by FACS analysis and by theinduction of the cell-cycle regulated gene MCM2(Figure 5D). Although these data confirmed thatERG1.2 regulation was associated to cell-cycle pro-gression, preliminary experiments failed to showanydirect effect of ERG1.2 overexpression on cell pro-liferation (data not shown).

Regulation of ERG1.2 Protein Expression

Tocharacterizebetter theexpressionof ERG1.2,wedeveloped a polyclonal antibody directed againstthe bacterially expressed carboxyterminal part ofERG1.2. The affinity-purified 2B#8 antiserum speci-fically recognized the ERG1.2 fragment overex-pressed by the e construct in NIH3T3 cells (datanot shown). The same antibody recognized the ex-pression of an approximately 35-kDa protein co-migrating with the transfected STAG1a protein inuntransfected C3H10T1/2 fibroblasts (Figure 6A).Upon treatment with EGF, IGF, or PDGF, its expres-sion increased after 1.5 h, remained high until 9 h,and declined below detectability after 24 h, confirm-

ing also at the protein level that ERG1.2 was growthfactor–regulated. Similar to TC-1S, NIH3T3, andBalb-3T3 cells, ERG1.2 transcript was cell-cycle–regulated in these cells (Figure 6B). When C3H10T1/2 cells were serum-starved, they stopped growing(data not shown) and the expression of the 35-kDaERG1.2 declined to undetectable levels. Upon FBSstimulation, ERG1.2 protein was upregulated after4 h of treatment and peaked after 12 h, beforedeclining to the level of exponentially growing cells.These results therefore confirmed that ERG1.2 wasgrowth factor– and cell-cycle–regulated at theprotein level also.Despite the large amount of mRNA expressed in

those cells, the initial attempt to characterize ERG1.2protein expression in NIH3T3 and TC-1S cells result-ed in the detection of a very faint band whose accu-mulation was slightly increased by EGF (data notshown). The interaction we demonstrated betweenERG1.2 and NEDD4 suggested that ERG1.2 proteinexpression could be controlled through a protea-some-mediated degradation. Indeed, the protea-some inhibitor MG132 induced the accumulationof the transfected, the constitutive endogenous, andthe EGF-induced ERG1.2 protein in NIH3T3 cells(data not shown). MG132 also induced accumula-tion of the endogenous ERG1.2 protein inC3H10T1/2 cells (Figure 6A), confirming that its expressionwasat least partially controlled through the proteasomedestruction pathway.Interestingly, the 2B#8 antibody also recognized

the human STAG1. This allowed us to show thatSTAG1 low constitutive expression could be stronglyinduced upon treatment with either EGF or MG132in Ha-Cat cheratinocytes and the two substances didnot appear additive (Figure 6C). Together with pre-vious observations, these data indicate that STAG1proteinmay also be regulated at post-transcriptionallevels (i.e., protein stability) by EGF in human cellsalso.

STAG1 Is Overexpressed in Human Breast Cancer

With the exception of Jurkatt cells, we detectedexpression of at least one STAG1 isoform in all celllines we tested (Table 2), confirming the broadexpression of this gene. Although EGF modulatesSTAG1 expression in a number of human cell lines(Figure 3 andTable 2), we noticed a high constitutiveand EGF-insensitive expression of the STAG1 andERG1.2 transcripts in many human and mousecancer cell lines, including the murine mammarycancer cell line TSA (Tables 1 and 2). In particular, weobserved high levels of expression in many breastand ovarian cancer cell lines (Figure 7A and Table 2),whereas we detected lower levels in the non-cancerbreast cell line MCF10A. This observation suggeststhat STAG1might be constitutively overexpressed inbreast and ovarian cancer cells compared to theirnormal counterparts. Indeed, we found increased

Figure 6. ERG1.2 protein expression is growth factor and cell-cycle regulated. (A) Western blot analysis of ERG1.2 proteinexpression in C3H10T1/2 cells stimulated with the described growthfactors. The 2B#8 antiserum recognized a growth factor regulatedprotein co-migrating with the overexpressed STAG1a. The protea-some inhibitor MG132 induced accumulation of the same protein,confirming that ERG1.2 was degraded by the proteasome pathway.The immunostaining with the a-tubulin antibody provided a loadingcontrol. In the last two lanes a small amount (1/10 the content of theother lanes) of cell extract from STAG1a and STAG1b transfectedcells was loaded. (B) Following serum starvation in 0.5% FBS for 24 h,C3H10T1/2 cells restimulated with 10% FBS for the indicated time.RNA and protein extracts were then analyzed. Ethidium bromidestaining of the 28S RNA and actin immunostaining were providedas loading controls for Northern and Western blotting, respectively.(C) The 2B#8 antiserum also recognized an EGF- and MG132-regulated protein in human Ha-Cat cells.

STAG1/PMEPA1/ERG1.2 IN BREAST AND OVARIAN CANCER 195

STAG1 expression in three of three primary breastcancer cases compared to controls (Figure 7C). Inter-estingly, although we could not detect any furtherincrease upon EGF treatment in most breast andovarian cancer cell lines, STAG1 expression couldbe regulated by EGF in the breast cell line MCF10A,but also in SK-BR-3 and ZR-75-1 breast cancerand OVCAR-3 ovarian cancer cells (Figure 7A and

Table 2). Western blot analysis confirmed a highexpression of STAG1 protein in BT474, BT20, andINTOV-1 cell lines (Figure 7B), which also showedvery high mRNA expression. Lower levels were alsoobserved in MDA-231 and SKOV-3 (not shown).Again, in all these cell lines, we did not detect anyfurther induction upon EGF addition (Figure 7B anddata not shown). Lower and partially inducible

Figure 7. STAG1 is overexpressed in breast and ovarian cancer.(A) STAG1 mRNA expression was studied by Northern blotting (A, C,D) and Western blotting (B) in the reported breast and ovarian cancercell lines (A, B, D), and in three primary breast cancers (C, lane 1, 2,3), a pool of normal breast donors (C, lane 4) and a normal lung (C,

lane 5). The presence of lung tissue in panel C allowed comparisonwith Figure 3B. (D) Both AG1478 and PD168393 EGF receptorinhibitors reduced STAG1 expression in the breast cancer and EGFR-amplified BT-20 cell line.

Table 1. Expression of ERG1.2 Isoforms in Mouse Cell Lines and Regulation by EGF

Cells

2.4 kb Form 5 kb Form 1.9 kb Form

CTR EGF CTR EGF CTR EGF

NeuroendocrineTC-1S þ þþþþ þ þþ þ þþPC-12 � þ þ þ þ þB104 þþþ þþþ þþþ þþþ þþþ þþþN18 � � þ þ þ þ

EpithelialTS-Aa þþþ þþþþ þ þþ þ þþMv1Lub þ ND þþþ ND þ ND

FibroblastNIH-3T3 þ þþþþ þ þþ þ þþBALB-3T3 þ þþþþ þ þþ þ þþC3H10T1/2 þþ þþþ þ þþ þ þþ

MuscleC2C12 þþ ND þþ ND þþ ND

EmatopoieticFLCc � � � � � �

aMammary gland tumor cell line.bMink cell line.cFriend leukemia cells.

196 GIANNINI ET AL.

expression was also detected in MDA-175 andOVCAR-3 (Figure 7B). These data suggested that theconstitutive activation of the EGF receptor pathwaytypically present inmany breast and ovarian cancersmight be responsible for the high constitutiveexpression of STAG1. Indeed, treatment of theEGFR-amplified BT-20 mammary cancer cells withthe EGFR inhibitors AG1478 and PD168393 reducedSTAG1 expression (Figure 7D), indicating that theconstitutive activation of the EGF receptor pathwaymight be at least partially responsible for STAG1overexpression in breast cancer cell lines.

STAG1 Is Amplified in Human Breast Cancer Cell Linesand Amplified and Rearranged in One Primary

Ovarian Cancer Case

The 20q13.3 chromosomal region, where STAG1maps, is frequently involved in amplification andchromosomal rearrangements in many human neo-plasias, including breast cancer. Thus we investi-

gated whether the STAG1 locus was also involved insuch processes in human breast cancer cell lines andprimary tumors. Southern blot analysis of 10 breastcancer cell lines with different probes revealed theamplification of the entire STAG1 locus in the BT-474 cell line (Figure 8A). Furthermore, although wefailed to reveal gross genetic abnormalities of STAG1locus in 15 primary breast cancers, we found thepresence of an amplified and rearranged fragment in1 (case #79) of 15 primary ovarian cancers (Figure 8B,line 5). The amplified and rearranged bands were in-dependently identified on genomic DNA separatelydigested with either EcoRI or HindIII (Figure 8C).

DISCUSSION

ERG1.2 Belongs to a New Gene Family

Through a differential display approach we clon-ed ERG1.2, a new EGF inducible gene. The highhomology between the mouse and human cDNA

Table 2. Expression of STAG1 Isoforms in Human Cell Lines and Regulation by EGF

Cells

2.4 kb Form 5 kb Form 1.9 kb Form

CTR EGF CTR EGF CTR EGF

NeuroendocrineSK-N-BE þ ND þ ND � NDSH-EP þ ND þ ND � NDKCNR þþ ND þþþ ND � ND

EpithelialHaCat þþ þþþþ þþ þþþ þþ þþþA549 þþþþ þþþþ þþþ-

þþþþþ-

þþþþþ þþþ

HeLa þþþþ ND þþþ ND þþ NDMCF10A* þ þþ þ þþ þ/� þ/�PMC-42{ þ/� ND þ/� ND � NDBT-549{ þ ND þ ND � NDMCF-7{ þþþ þþþ þþþ þþþ þ þMDA-175{ þþ þþ þþ þþ þ þZR-75-1{ þ þþþ þ þþþ þ þþMDA-231{ þþþþ þþþþ þþþþ þþþþ þþ þþBT-474{ þþþþ þþþþ þþþþ þþþþ þþ þþBT-20{ þþþþ þþþþ þþþþ þþþþ þþ þþSK-BR-3{ þ þþ þ þþ þþ þþMDA-435{ þþþ þþþ þþþ þþþ þ þIGROV-1{ � � � � � �INTOV-1{ þþþþ þþþþ þþ þþ þþþþ þþþþSKOV-3{ þþþ þþþ þþ þþ þ þOVCAR-3{ þ þþ � � � þ

FibroblastWI38 þ þþ þ þþ þ þþ

SarcomaSaos-2 þ þþþ þ þþ þ/� þ/�RD þþ ND þþ ND þþ ND

EmatopoieticJurkatt � � � � � �

*Mammary gland cell line.{Breast cancer cell line.{Ovarian cancer cell line.

STAG1/PMEPA1/ERG1.2 IN BREAST AND OVARIAN CANCER 197

sequence, ERG1.2 mapping onto mouse chromo-some 2 in a region synthenic to human chromosome20q13.3 and the very similar genomic architecture ofthe two genes indicate that ERG1.2 is the murineortholog of the STAG1 gene. Both in humans and inmice, two highly homologous genes exist (calledC18orf1 and STAG1 in humans and mC18orf1 andERG1.2 in mice) and can give origin to multipletranscripts and protein isoforms by different mecha-nisms, including alternative splicing and multipletranslation starts. All the proteins share themodularstructure of potential transmembrane receptors.Indeed, a type 1b transmembrane domain is con-served in all members. C18orf1a has an LDLRAdomain preceding its transmembrane domain. Thiselement, which is typically repeated several times inthe large N-terminus of the LDL receptors and LDLreceptor related proteins (LRPs), is involved in theinteractionwithLDLand calcium [18,19].Whether asingle LDLRA domain may confer ligand-bindingspecificity to C18orf1a is presently unknown. Inter-estingly, we have shown that a heavily glycosylat-ed domain substitute for the LDLRA in STAG1b/PMEPA1/ERG1.2. Therefore, both C18orf1a andSTAG1b/PMEPA1/ERG1.2 have in principle thepossibility of functioning as receptors with poten-tially different ligand specificity. Having no ligand-binding region, STAG1a andC18orf1bmight insteadbe involved in modulating the activity of the longerisoforms or work as decoy receptors.Through the very conserved proline-reach carbox-

yterminus containing several SH3- andWW-bindingdomains, STAG1b (and potentially its homologs)interacts with cellular proteins containing WW orSH3 motifs, such as NEDD4, the Yes-associatedprotein YAP65 and GRB-2. By this means, this newprotein family could be involved in a number ofregulatory and signaling pathways. Through its

interaction with NEDD4, a ubiquitin-protein ligasethat regulates the stability and function of mem-brane proteins [20], ERG1.2/STAG1 is probablytargeted for degradation through the proteasomepathway. Indeed, the proteasome inhibitor MG132increases ERG1.2 and STAG1 protein stability.ERG1.2/STAG1 interaction with the ubiquitousphosphoprotein YAP65 is of difficult interpretationat themoment, since its biological functions are stilllargely obscure. However, because of its modularstructure and based on its interactions with nuclearproteins such as p73 and p53BP2 [17,21], YAP65 isbelieved to work as an adaptor molecule linkingextracellular signaling events to nuclear functions.Thus we hypothesize that it might also connectERG1.2/STAG1 signaling to nuclear functions. IncontrastGRB2 is awell-knownadaptermolecule thatlinks growth factor receptors to MAPK signaling andthus to the regulation of a number of cellularfunctions, including growth control and stressresponses [22,23]. Thus ERG1.2 might also take partin such processes through its interaction with GRB2.

ERG1.2 Is a Tightly Controlled Growth Factor–

and Cell-Cycle–Regulated Gene

In cultured cells, ERG1.2 expression largelyincreases above the constitutive level upon stimula-tion with growth factors; on the contrary, it isabolished in quiescent cells. Stimulation of quies-cent cells to enter the cell cycle is accompanied by asuddenpeak inERG1.2 expression that clearly occursinG1 and, before cells enter S phase, quickly declinesto the levels that are usually found in asynchro-nously cycling cells. Addition of growth factors toquiescent Balb3T3 cells is not sufficient to induceERG1.2 expression per se, unless it is fully competentto drive cells into the cell cycle, thus suggesting thatERG1.2 induction is strictly linked to cell-cycleprogression. Transient ERG1.2 overexpression inseveral cell types failed to demonstrate a sharp effecton cell proliferation and/or apoptosis. Thus furtherexperiments are needed to clarify whether ERG1.2expression is indeed required for cell-cycle progres-sion. In any case, ERG1.2 expression appears tightlycontrolled. Its inductionby growth factors is likely tooperate largely at the transcriptional level, since it isCHX resistant. However, oncemade, ERG1.2 proteincan be degraded through the proteasome pathway.We have evidence that a 50-UTR containing ERG1.2construct is poorly translated in transfected cells andalso ERG1.2 endogenous message is poorly asso-ciated with heavy polysomes, suggesting a strongcontrol at the level of mRNA translation (Cardinaliand Giannini, unpublished). Consistent with thesedata, most cells we examined only expressed verylittle ERG1.2 protein. In contrast, breast and ovariancancer cells have an increased expression of ERG1.2at themRNAandprotein level. It is possible thathighlevels of ERG1.2 could be detrimental for untrans-

Figure 8. STAG1 locus is directly targeted by genetic events inbreast and ovarian cancer. Genomic DNA extracted from theindicated breast cancer cell lines (A) and primary ovarian cancers(B), were digested with HindIII, BamHI, and EcoRI separately (A) orwith EcoRI and HindIII together (B) and screened by Southern blotwith a probe located on exon 4 of STAG1. (C) Case #79 in lane panelB, lane 5 was then separately digested with EcoRI and HindIII andhybridized as before. A normal DNA (5 times more abundant) wasincluded as control. Asterisks indicate the expected bands ascompared with the pattern obtained for the control DNA. Arrow-heads indicated the amplified and rearranged fragments.

198 GIANNINI ET AL.

formed cells, in which its expression is usually keptlow through poor translation and active degradationthrough the proteasome pathway. Interestingly,cancer cells may have acquired the possibility totolerate or even gain advantages by higher levels ofERG1.2 expression.

ERG1.2/STAG1 Is Overexpressed and Amplified in

Breast/Ovarian Cancer

STAG1 is overexpressed in renal cell and colorectalcancer [14]. Our data now clearly indicate thatSTAG1 is also overexpressed in breast and ovariancancer cell lines and primary tumors. Thus, accord-ing to its name, STAG1 overexpression appears to bea common feature ofmany solid tumors. Althoughatthe moment we lack evidence for its biological role,we know more about the mechanisms that maycontribute to its deregulation in cancer. Of course,epigenetic eventsmight contribute to the regulationof STAG1 promoter. However, it is likely that inmany instances STAG1 overexpression might be theconsequence of the constitutive activation of agrowth factor dependent pathway. In particular, wenoticed that, although the low constitutive expres-sion of STAG1/ERG1.2 is readily upregulated by EGFin both human and mouse non-cancer cell lines, itsexpression was rather high and no longer responsiveto the growth factor in many cancer cells. Theabnormal activation of the EGF pathways is acommon theme in epithelial cancer [1]. In 20–30%of the breast cancer cases, it involves the over-expression of ErbB2 because of genetic amplificationand/or the overexpression of a normal or aberrantEGFR [1,3,4]. Thus it is likely that STAG1/ERG1.2overexpression in many breast cancer cell lines andprimary tumors may be a consequence of theconstitutive activation of the EGF pathway. Indeed,we have shown that chemical inhibition of the EGFRpathway impairs STAG1 expression in EGFR ampli-fied breast cancer cells. Xu et al. [15] have shown thatSTAG1/PMEPA1 is highly expressed in prostatecancer cells and is inducible by the hormone inandrogen dependent mouse xenograft, whereas it isconstitutively expressed in androgen-independenttumors. In this respect, it seems rather interesting tous that in androgen-independent advanced prostatecancers TGFa sustains an autocrine EGFR signalingactivation [6] and that the expression of EGFRcorrelates with the progression of prostate cancertowards androgen-independence [24]. Thus the in-creased expression of STAG1 in androgen refractoryprostate cancer xenografts [15] also might be due tothe constitutive activation of the EGF pathway.STAG1 is genetically amplified inonebreast cancer

cell line and is amplified and rearranged in one of the15 primary ovarian cancers that we screened. There-fore, it is likely that genetic amplification may alsocontribute to determining STAG1 overexpressionin human cancers. Gain of chromosome 20q is a

frequent genetic abnormality in many cancers,including breast, gastrointestinal, prostate, bladder,pancreatic cancer, and melanomas, and may thusbe responsible for the expression of extra-copies ofthe STAG1 gene. Finally, our data also provided apreliminary but substantial evidence that STAG1locus was directly targeted by a genetic rearrange-ment in one primary ovarian cancer, suggesting thatSTAG1 is not only a bystander within the geneticregion of relevance for cancer development.In conclusion, we showed that ERG1.2/STAG1 is a

cell-cycle–regulated gene whose expression is fre-quently reactivated in human tumors via directgenetic targeting or indirect mechanisms andmightbe regarded as a potential tumormarker of biologicalrelevance for the carcinogenetic process.

ACKNOWLEDGMENTS

We thankDr. M. Sudol for providing theWW-GSTconstructs and Dr. O. Segatto for the SH3-GSTconstructs. This work was supported by grants fromAssociazione Italiana per la Ricerca sul Cancro,Ministry of Health, the National Research Council(CNR), the Ministry of University and Research, thePasteur Institute, Cenci-Bolognetti Foundation.

REFERENCES

1. Yarden Y. The EGFR family and its ligands in human cancer.Signalling mechanisms and therapeutic opportunities. EurJ Cancer 2001;37(Suppl 4):S3–S8.

2. Slamon DJ, Clark GM, Wong SG, Levin WJ, Ullrich A,McGuire WL. Human breast cancer: Correlation of relapseand survival with amplification of the HER-2/neu oncogene.Science 1987;235:177–182.

3. Wikstrand CJ, Reist CJ, Archer GE, Zalutsky MR, Bigner DD.The class III variant of the epidermal growth factor receptor(EGFRvIII): Characterization and utilization as an immu-notherapeutic target. J Neurovirol 1998;4:148–158.

4. Moscatello DK, Holgado-Madruga M, Godwin AK, et al.Frequent expression of a mutant epidermal growth factorreceptor in multiple human tumors. Cancer Res 1995;55:5536–5539.

5. Salomon DS, Brandt R, Ciardiello F, Normanno N. Epidermalgrowth factor-related peptides and their receptors in humanmalignancies. Crit Rev Oncol Hematol 1995;19:183–232.

6. Scher HI, Sarkis A, Reuter V, et al. Changing pattern ofexpression of the epidermal growth factor receptor andtransforming growth factor alpha in the progression ofprostatic neoplasms. Clin Cancer Res 1995;1:545–550.

7. Screpanti I, Scarpa S, Meco D, et al. Epidermal growth factorpromotes a neural phenotype in thymic epithelial cells andenhances neuropoietic cytokine expression. J Cell Biol 1995;130:183–192.

8. Screpanti I, Meco D, Scarpa S, et al. Neuromodulatory loopmediated by nerve growth factor and interleukin 6 in thymicstromal cell cultures. Proc Natl Acad Sci USA 1992;89:3209–3212.

9. Vacca A, Di Marcotullio L, Giannini G, et al. Thrombospon-din-1 is a mediator of the neurotypic differentiation inducedby EGF in thymic epithelial cells. Exp Cell Res 1999;248:79–86.

10. Giannini G, Alesse E, Di Marcotullio L, et al. EGF regulates acomplex pattern of gene expression and represses smoothmuscle differentiation during the neurotypic conversion ofthe neural-crest-derived TC-1S cell line. Exp Cell Res 2001;264:353–362.

STAG1/PMEPA1/ERG1.2 IN BREAST AND OVARIAN CANCER 199

11. Harlow E, Lane D. Antibodies, a laboratory manual. ColdSpring Harbor, NY: Cold Spring Harbor Laboratory. 1988.

12. Chen CY, Shyu AB. AU-rich elements: Characterization andimportance in mRNA degradation. Trends Biochem Sci 1995;20:465–470.

13. Freter RR, Irminger JC, Porter JA, Jones SD, Stiles CD. A novel7-nucleotide motif located in 30 untranslated sequences ofthe immediate-early gene set mediates platelet-derivedgrowth factor induction of the JE gene. Mol Cell Biol 1992;12:5288–5300.

14. Rae FK, Hooper JD, Nicol DL, Clements JA. Characterizationof a novel gene, STAG1/PMEPA1, upregulated in renal cellcarcinoma and other solid tumors. Mol Carcinog 2001;32:44–53.

15. Xu LL, Shanmugam N, Segawa T, et al. A novel androgen-regulated gene, PMEPA1, located on chromosome 20q13exhibits high level expression in prostate. Genomics 2000;66:257–263.

16. Yoshikawa T, Sanders AR, Esterling LE, Detera-Wadleigh SD.Multiple transcriptional variants and RNA editing inC18orf1,a novel gene with LDLRA and transmembrane domains on18p11.2. Genomics 1998;47:246–257.

17. Espanel X, Sudol M. Yes-associated protein and p53-bindingprotein-2 interact through their WW and SH3 domains. J BiolChem 2001;276:14514–14523.

18. van Driel IR, Goldstein JL, Sudhof TC, Brown MS. Firstcysteine-rich repeat in ligand-binding domain of low den-sity lipoprotein receptor binds Ca2þ and monoclonal anti-bodies, but not lipoproteins. J Biol Chem 1987;262:17443–17449.

19. Mahley RW. Apolipoprotein E: Cholesterol transport proteinwith expanding role in cell biology. Science 1988;240:622–630.

20. Harvey KF, Kumar S. Nedd4-like proteins: An emergingfamily of ubiquitin-protein ligases implicated in diversecellular functions. Trends Cell Biol 1999;9:166–169.

21. Strano S, Munarriz E, Rossi M, et al. Physical interaction withYes-associated protein enhances p73 transcriptional activity.J Biol Chem 2001;276:15164–15173.

22. Tari AM, Lopez-Berestein G. GRB2: A pivotal protein in signaltransduction. Semin Oncol 2001;28:142–147.

23. Kyriakis JM, Avruch J. Mammalian mitogen-activatedprotein kinase signal transduction pathways activatedby stress and inflammation. Physiol Rev 2001;81:807–869.

24. Di Lorenzo G, Tortora G, D’Armiento FP, et al. Expression ofepidermal growth factor receptor correlates with diseaserelapse and progression to androgen-independence inhuman prostate cancer. Clin Cancer Res 2002;8:3438–3444.

200 GIANNINI ET AL.