Pleiotrophin Stimulates Fibroblasts and Endothelial and Epithelial Cells and Is Expressed in Human Cancer*

Previously we reported the purification of the hep- arin-binding growth factor pleiotrophin (PTN) from supernatants of the human breast cancer cell line MDA-MB-23 1. To investigate further the biological activities of PTN and its potential role in cancer, we cloned a PTN cDNA and expressed the gene in a human kidney and in a human adrenal carcinoma cell line (SW-13). The supernatants harvested from cells trans- fected with PTN contained a heparin-binding specific protein of an apparent molecular mass of 18 kDa. These supernatants stimulated the proliferation of endothelial cells as well as the anchorage-independent growth of SW-13 cells and of normal rat kidney fibroblasts. Furthermore, SW-13 cells transfected with PTN acquired autonomous growth in soft agar and were tumorigenic in athymic nude mice. In contrast to these results with PTN from human cells, PTN ob- tained from insect cells (Sf9) using recombinant baculovirus as a vector was biologically inactive. We detected high levels of PTN mRNA in 16 of 27 primary human breast cancer samples (62%) as well as in 8 of 8 carcinogen-induced rat mammary tumors. Further- more, 9 of 34 human tumor cell lines of different origin showed detectable PTN mRNA. We conclude that PTN may function as a tumor growth and angiogenesis factor in addition to its role during embryonic development.

Previously we reported the purification of the heparin-binding growth factor pleiotrophin (PTN) from supernatants of the human breast cancer cell line MDA-MB-23 1. To investigate further the biological activities of PTN and its potential role in cancer, we cloned a PTN cDNA and expressed the gene in a human kidney and in a human adrenal carcinoma cell line (SW-13). The supernatants harvested from cells transfected with PTN contained a heparin-binding specific protein of an apparent molecular mass of 18 kDa. These supernatants stimulated the proliferation of endothelial cells as well as the anchorage-independent growth of SW-13 cells and of normal rat kidney fibroblasts. Furthermore, SW-13 cells transfected with PTN acquired autonomous growth in soft agar and were tumorigenic in athymic nude mice. In contrast to these results with PTN from human cells, PTN obtained from insect cells (Sf9) using recombinant baculovirus as a vector was biologically inactive. We detected high levels of PTN mRNA in 16 of 27 primary human breast cancer samples (62%) as well as in 8 of 8 carcinogen-induced rat mammary tumors. Furthermore, 9 of 34 human tumor cell lines of different origin showed detectable PTN mRNA. We conclude that PTN may function as a tumor growth and angiogenesis factor in addition to its role during embryonic development.
Polypeptide growth factors have been shown to play important physiological roles in the timely development of tissues during embryonal and neonatal growth, and their expression is tightly regulated. On the other hand, polypeptide growth factor gene expression is deregulated in tumor cell lines as well as in solid tumors, and the activity of polypeptide growth factors appears to contribute significantly to autocrine and paracrine stimuli (for a review see Refs. 1 and 2).
Previously we have reported the purification of an 18-kDa heparin-binding growth factor from the conditioned medium of a human breast cancer cell line MDA-MB-231 (3). The NH2-terminal sequence of this factor did not bear homology to any known heparin-binding FGFs but was homologous to a developmentally regulated protein which has recently been * This work was supported in part by National Cancer Institute Grant UO1 CA51908 and by a grant from the Deutsche Forschungsgemeinschaft (Germany) (to A. W. and N. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ll To whom correspondence should be addressed Georgetown University, Lombardi Cancer Research Center, 3800 Reservoir Rd. NW, Washington, D. C. 20007. Tel.: 202-687-3672;Fax: 202-687-7505. described by numerous laboratories under different names such as "heparin-affin regulatory protein (HARP)" (4, 5), heparin-binding neurotrophic factor (HBNF) (6,7) or p18 (8) from bovine brain, heparin-binding growth associated molecule (HB-GAM) from perinatal rat brain (9)(10)(11), heparinbinding growth factor 8 ( E ) , osteoblast-specific factor from mouse brain (13), and pleiotrophin (PTN)' from human placenta or rat brain (11). Due to its numerous biological activities, we have decided to use the latter name for this growth factor. PTN belongs to a novel family of heparin-binding proteins which include the structurally related midkine proteins and appear to function during brain development (for a review see Ref. 14). The mitogenic activity of PTN, however, is still controversial (14). Several laboratories have described mitogenic activity of PTN purified from different sources for endothelial cells (4,5,7) and fibroblasts (6,7,11,12). We reported that a purified preparation of PTN stimulates colony formation in soft agar of the epithelial cell line SW-13, and we identified PTN in this preparation by protein sequencing (3). However, other investigators have disputed an intrinsic growth factor activity of PTN and have attributed the activities to FGFs or other growth factors contaminating the respective preparations (15)(16)(17).
Our present study addresses this controversy by expressing the wild-type PTN cDNA under the control of a strong CMV promoter in two human cell lines. A PTN point mutant with a premature translation stop codon served as a negative control in these experiments. We show that supernatants from cells transfected with the wild-type construct but not the point mutant contain an 18-kDa heparin-binding protein that is immunoreactive and stimulates endothelial and SW-13 epithelial cells as well as fibroblasts. Furthermore, after expression of wild-type PTN, SW-13 cells acquired autonomous growth in soft agar and grew into tumors in athymic nude mice. This provides additional independent evidence that the PTN gene product can act as a growth factor in tumors. Furthermore, in agreement with a recent report from other authors (16), we were unable to obtain PTN in a biologically active form from insect cells which overproduced the protein after infection with PTN-recombinant baculovi- The potential significance of PTN for tumor growth is addressed in further experiments studying expression of this gene in primary tumors and in tumor cell lines. About onefourth of a series of established human tumor cell lines showed rUS.

25889
expression of PTN as assessed by RNase protection assays. Most significantly, more than half of a set of 27 primary human breast cancer samples contained high levels of P T N mRNA. Finally, carcinogen-induced rat mammary tumors also scored positive for P T N mRNA.
We conclude that the protein product of this developmentally active gene seems to play a significant role in the progression of human tumors. Based on its mitogenic activities it most likely functions as a tumor angiogenesis factor.

MATERIALS AND METHODS
Cell Lines and Tissues-BT-474, BT-579, Hs-578T, MCF-7, MDA-453, MDA-MB-468, SK-BR-3, T47D, ZR-75-1 (human breast cancer), DU-145, LNCaP, PC-3 (human prostate cancer), A-549 (human lung cancer), OVCAR-3, PA-I (human ovarian cancer), A-204 (human rhabdomyosarcoma), A-431 (human epidermoid carcinoma), HL-60 (human promyelocytic leukemia), SW-13 (human adrenal cancer), HGT-1, KATOIII (human stomach cancer), 293 (human embryonal kidney), FBHE (fetal bovine heart endothelium), NRK (normal rat kidney fibroblasts clone 49F) were obtained from the American Type Culture Collection (Rockville, MD). The following cell lines or sublines were kindly made available by colleagues at the Lombardi Cancer center: MCF/LY2 (human breast cancer) by Dr. M. E. Lippman (18) Tumor tissues were immediately snap-frozen in liquid nitrogen before being stored at -70 "C. Generation of the PTN Expression Vector for Human Cell Lines-The PTN-gene was cloned from MDA-MB-231 human breast cancer cells from which the protein had been purified (3). Poly(A+) RNA was prepared from approximately 1 X 10' cells. After lysis of the cells in 5 M guanidinium isothiocyanate, 0.05 M Tris-C1 (pH 7.5), 5% pmercaptoethanol, 10 mM sodium EDTA, poly(A+) RNA was isolated by oligo(dT) affinity chromatography (Invitrogen Corp., San Diego, CA). Poly(A+) RNA was then reverse-transcribed to cDNA, and subsequently a specific fragment was amplified by the polymerase chain reaction (PCR) using the Gene AmpTM RNA reagants (Perkin-Elmer Cetus). Sense and antisense primers corresponding to the 5'and 3'-untranslated regions of the human PTN transcript were synthesized (see Fig. lB), with additional HindIII and XbaI sites to facilitate subsequent cloning of amplified PCR products. Three extra nucleotides (GGT) were included at both ends of the primers to improve recognition of the cleavage sites by the restriction enzymes. The primers were 5"GGTTCTAGAT ATGTTCCACA GGTGA-CATC-3' (3"antisense) and 5"GGTAAGCTTA GAGGACGTTT CCAACTCAA-3' (5"sense). An Eppendorf Microcycler E was used for the PCR reaction (initially 2 min at 95 "C, then 1 min at 95 "C and 1 min at 50 "C for 35 cycles, and finally 7 min at 60 "C for 1 cycle). The PCR product was purified using the ion exchange columns and reagants provided by Qiagen (Qiagen Inc., Chatsworth, CAI. The product of the PCR reaction was digested with HindIII and XbaI and cloned into the pRc/CMV expression vector (see Fig. U

t ; Invitrogen).
This plasmid preparation was amplified in bacteria (strain DH5a). Positive clones were picked and the inserts were sequenced by the dideoxy chain termination method (18) using Sequenase version 2.0 from United States Biochemical Corp.
Transient and Stable Transfections of Human Cells-The transfections were carried out using the calcium phosphate precipitation/low COz method as described (19). Briefly, approximately 5 X lo5 cells/ IO-cm plate were seeded and incubated overnight in growth medium. Then plasmid DNA (25 pg) was mixed with 0.5 ml of 0.25 M CaCL MB-134, MDA-MB-231, MDA-MB-361, MDA-MB-435, MDA-MB-and 0.5 ml of 2 X BES and incubated for 20 min at room temperature. The calcium phosphate DNA solution (1 ml) was then added to cells in 9 ml of growth medium and incubated for 24 h at 35 "C under 2.7% COZ. The cells were rinsed twice with growth medium, refed with growth medium, and incubated for 24 h at 37 "C under 5% COz. Stable transfectants were selected by culturing cells in the presence of G418 (250 pg/ml; GIBCO/BRL). Clonal cell lines were obtained by serial dilution. Cells were diluted to approximately one ce11/100 p1 of media and plated at 100 pl into 96-well plates. After 2 weeks, the wells showing growth of an individual clone were harvested and propagated.
Overexpression of Secreted PTN in Sf9 Insect Cells Using the Baculouirus System-PTN was overexpressed using a recently developed co-transfection system of a replication-deficient baculovirus in combination with a plasmid containing complementing genes (Pharmingen, San Diego, CA) and the PTN insert under the polyhedrin promoter. The PTN insert from the pRc/CMV described above was subcloned into pVL1393 (Pharmingen). pVL1393 contains the necessary genes to rescue the replication-deficient baculovirus in insect cells. Plasmid and baculovirus DNA were co-transfected into Sf9 cells (ATCC) kept at 27 "C in Sf-900 media (GIBCO/BRL). After generating recombinant virus with the PTN-gene insert and infection of Sf9 cells, approximately 10 mg of recombinant PTN were obtained from cell supernatants as assessed after NHz-terminal sequencing of an aliquot of this preparation (see "Results").
Assays for PTN Secreted from PTN-expressing Cells-Media conditioned by different transfected cells were collected and loaded onto heparin-Sepharose columns (Pharmacia LKB Biotechnology Inc.) pre-equilibrated with phosphate-buffered saline. The columns were then washed with this buffer, and bound proteins were eluted either with a step-gradient of 0.4, 0.9, and 2.0 M NaCl in 10 mM Tris buffer (pH 7.5) or in bulk with the 2.0 M NaCl step (3).
Immunoassay-Proteins present in 10O-pI aliquots were adsorbed onto 96-well plates by overnight drying at 37 "C. The remaining free binding sites in the wells were then blocked with 1% bovine serum albumin dissolved in phosphate-buffered saline with 0.5% Tween 20 (PBST) for 2 h at room temperature and washed three times with this buffer. After incubation with a rabbit antiserum against a peptide comprising the 10 NHz-terminal amino acids of mature, secreted PTN (NHz-GKKEKPEKEK; see Ref. 3) for 1 h at 4 "C, the plate was washed again three times with PBST. The second antibody, peroxidase-labeled affinity-purified goat anti-rabbit IgG (Kirkegaard & Perry Laboratories, Inc., Gaithersburg, MD) was added to the plate at a dilution of 1/250 and incubated for 1 h at 4 "C. The plate was then washed four times with PBST and incubated with peroxidase substrate 2,2'-azino-bis(3-ethylbenz-thiazoline-6-sulfonic acid (Sigma) for 5-30 min. Absorbance was then measured using a microplate reader (Dynatech Laboratories Inc., Chantilly, VA) at 405 nm.
Metabolic Labeling-For in vivo metabolic radiolabeling, transiently transfected cells were washed once with cysteine-free growth medium (GIBCO/BRL) and then incubated with this medium and 100 pCi/ml of [35S]cysteine. After 24 h the cell supernatants were collected and immediately subjected to heparin-Sepharose chromatography on a 0.4-ml column as described above. An aliquot of the 2 M NaCl eluates was analyzed by SDS-PAGE and autoradiography as described (3,19).
Growth Assays-Studies of anchorage-independent growth of NRK fibroblasts and of SW-13 cells in soft agar were carried out as described (19). Briefly, 10,000cells in 0.35% agar (Bactoagar; GIBCO/ BRL) were layered on top of 1 ml of a solidified 0.6% agar layer in a 35-mm dish. Material to be tested was filter-sterilized, and 100-pl aliquots from the heparin-Sepharose were added with the 800-p1 top layer unless indicated otherwise. Growth media with 10% FBS were included in both layers. Colonies more than 60 pm in diameter were counted after 2-3 weeks of incubation using an image analyzer.
The mitogenic assays with endothelial cells used an approach described previously by Klagsbrun et al. (20). In brief, 50-p1 aliquots of the eluates from heparin-Sepharose were added to about 5,000 cells/well using 24-well plates. After 4 days of incubation, the cells were detached and counted with a particle counter.
Tumor Growth in Animals-Female athymic nude mice (NCr nu/ nu; Harlan Sprague-Dawley, Indianapolis, IN) were injected into a mammary fat pad with 2 X IO6 cells and observed for at least 8 weeks for tumor formation.
Detection of mRNA by Northern Blots-Total RNAs were isolated with the RNAzol'" B method using commercially available reagants and protocols (RNAzol'" B; Tel-Test Inc.; Friendswood, TX). 30 r g of total RNA were separated by electrophoresis in 1.6% formaldehydeagarose gel and then blotted onto nylon membranes (Schleicher and Schuell). The blots were prehybridized in 6 X SSC (= 0.9 M NaC1, 0.09 M sodium citrate (pH 7.0)), 0.5% (w/v) SDS, 5 X Denhardt's solution (= 0.1% (w/v) Ficoll, 0.1% (w/v) polyvinylpyrrolidine, 0.1% (w/v) bovine serum albumin) (GIBCO/BRL) for 4 h a t 68 "C and hybridized overnight a t 68 "C in hybridization solution (same composition as the prehybridization solution with addition of 0.01 M EDTA (pH 8.0)) (GIBCO/BRL) containing a [a-32P]dCTP-labeled P T N cDNA probe. This probe was prepared by random primed DNA labeling (Boehringer Mannheim). After hybridization, blots were washed once with 1 X SSC and 0.1% SDS for 20 min, once with 0.2 X SSC and 0.1% SDS for 20 min at room temperature, and once for 20 min at 68 "C. Blots were then autoradiographed using intensifying screens a t -70 "C for 2 days.
Detection of mRNA by RNase Protection Assays-An [a-32P]UTPlabeled antisense riboprobe was prepared by in vitro transcription with SP6 RNA polymerase from linearized pPTNwt plasmid. Of this 832-hp probe, a 551-bp fragment will be protected by PTN mRNA present in the samples. As a loading control a 310-bp probe for the human acidic ribosomal phosphoprotein PO (3684) was used. A 210bp protected fragment is expected from this probe. 30 pg of total RNA were hybridized with 50,000 dpm of probe overnight a t 50 "C.
Samples were then digested with 40 pg/ml of RNase A (Boehringer Mannheim) for 30 min at 25 "C. The RNase digestion was terminated by adding 250 pg/ml of proteinase K (Boehringer Mannheim) and 0.5% SDS for 15 min a t 37 "C. The samples were extracted with phenol/chloroform once and precipitated together with 10 fig of tRNA in ethanol. Pellets were boiled in formamide loading buffer and separated on a 6% polyacrylamide sequencing gel. The gels were dried and exposed with intensifying screens at -70 "C for 2-4 days.

RESULTS
We previously purified an 18-kDa heparin-binding growth factor from the conditioned medium of an estrogen receptornegative human breast cancer cell line MDA-MB-231 (3). The NH,-terminal sequence of the purified protein was homologous to that of PTN purified by others from normal tissues (see Introduction), and an mRNA coding for PTN was expressed in the MDA-MB-231 cells. This purified protein was growth stimulatory in soft agar for the epithelial cell line SW-13 as well as for normal rat kidney fibroblast (NRK). To study the function of PTN and to provide evidence that growth stimulatory activities observed with our purified preparation were due to PTN, we decided to clone the cDNA for P T N from MDA-MB-231 cells, express the gene, and study the function of the recombinant product. We opted for expression in human cells, since PTN is secreted after cleavage of a 32-amino acid signal peptide (3) and contains 10 cysteine residues linked by disulfide bridges (8). Our earlier studies indicated that these disulfide links are required for biological activity since treatment with the reducing agent dithiothreitol destroys mitogenic activity of PTN (3). We also produced P T N on a large scale in insect cells using baculovirus as a vector. However, PTN produced in insect cells was biologically inactive in the growth assays (see below).

Constructs Used in the Expression Studies
For the studies on PTN function, we transfected into human cell lines a PTN construct with a strong constitutive CMV promoter. Our earlier studies with SW-13 cells and the K-FGF gene (19) and preliminary studies with the human kidney cells 293 had shown that this promoter supports high levels of gene expression in these cell lines. We utilized the commercially available pRc/CMV vector (Fig. 1A) which contains a CMV promoter upstream of a multiple cloning site and a transcription unit for the G-418 resistance gene for selection of stably transfected eukaryotic cells.
A data base search with the PTN cDNA showed that a stretch of the initial 97 nucleotides at the 5' end of PTN is 90.7% homologous to the antisense strand of the 3' end of human heat shock protein 70 cDNA (hsp7O) (GenBank/

FIG. 1. Construct used in the transfection studies (A) and comparison of wild-type and mutant PTN inserts ( B ) . A , the
plasmid map of the PTN cDNA construct used for transfection is shown. The original plasmid pRc/CMV was obtained from Invitrogen (San Diego, CA). The construction is described under "Materials and Methods." BGH PA, bovine growth hormone poly A site. B, the overall structure of the PTN cDNA is depicted at the top. The solid bar indicates the open reading frame (ORF) of PTN. Shaded bars in the 5'and 3"untranslated regions of P T N indicate areas that are homologous to regions in the antisense strands of the human cDNAs of hsp70 and L7, respectively (see "Results" for details). The start position of the primers used for PCR cDNA cloning is indicated by arrowheads, and the respective nucleotide sequence is given below. The position of the frame shift mutation in the ORF is symbolized by a vertical arrow, and the respective nucleotide and deduced amino acid sequence in the wild-type and mutant construct are shown. The full-length PTN cDNA and amino acid sequence are accessible through GenBank/EMBL with the accession numbers M57399 and P21246, respectively. EMBL accession number X04677). Furthermore, a stretch of 277 nucleotides at the very 3' end of the PTN cDNA is 98.6% homologous to the antisense 5' end of the human ribosomal protein L7 cDNA (GenBank/EMBL accession number X52967) (see Fig. 1B). To avoid interference of high levels of PTN gene expression with the expression of endogenous hsp70 or L7, we decided to include only the open reading frame (ORF) of PTN in our expression vector. We used PCR primers extending 22 nucleotides into the 5'-and 3"untranslated regions of PTN, respectively, to maintain the context of the natural translation initiation site of the PTN gene (see Fig. 1B). Additional Hind111 and XbaI sites were included at the ends of the primers to facilitate cloning of an amplified product.
A specific DNA product was PCR-amplified after reverse transcription of poly(A+) RNA from MDA-MB-231 cells as published previously (3). The resulting 569-bp product was Pleiotrophin Function and Expression cloned into pRc/CMV (Fig. 1A). Sequencing of the PCR product after bacterial amplification and selection of different bacterial clones showed that one construct, pPTNwt, contained the wild-type cDNA with the complete ORF. This ORF of pPTNwt contains 507 nucleotides and is identical to the reported human cDNA sequence of P T N (GenBank/EMBL accession number M57399). One of the bacterial clones picked contained a plasmid with a point mutation in the ORF of PTN. At position +130 (relative to the translation initiation site) an additional A was inserted (see Fig. 1B). This frame shift mutation results in a premature translation stop codon after amino acid 15 of the mature protein. The protein product expected from this frame shift construct is the signal peptide and a 15-mer peptide comprising 11 residues from the NH2 terminus of the mature PTN protein. We used this mutant construct as our negative control in further studies due to the fact that wild-type and mutant expression vectors only differ by a single nucleotide.

Demonstration of Secreted and Biologically Active P T N Protein in Transient Expression Experiments in Human Cells
In one series of experiments we transfected the human embryonal kidney cell line 293 with the wild-type (pPTNwt) and mutant (pPTNmu) constructs described above. CM from transiently transfected 293 cells were used for further analysis of secreted protein products. After in vivo metabolic [35S] cysteine labeling of transfected cells for 24 h, the cell supernatants were collected and passed through heparin-Sepharose columns. The eluted proteins from the columns were separated by SDS-PAGE (Fig. 2.4). A labeled protein migrating at about 18 kDa was the only species in this heparin affinitypurified material that is present in supernatants from pPTNwt but not from pPTNmu-transfected cells. The apparent molecular mass of this metabolically labeled protein was identical to that of the PTN protein originally purified from supernatants of the MDA-MB-231 breast cancer cells (3). This protein was also recognized in a Western blot using a rabbit polyclonal antiserum against PTN (data not shown). Furthermore, as shown in Fig. 2B, a high amount of P T N immunoreactivity was detected in the CM of pPTNwt-transfected 293 cells and no signal above background with CM from the pPTNmu controls.
Heparin affinity-purified CM collected from the transient expression experiment with 293 cells was also tested for growth-stimulating activity in SW-13, NRK, and endothelial cells. Fig. 2C shows that the CM from pPTNwt relative to pPTNmu-transfected cells stimulated the colony formation of SW-13 and of NRK cells in soft agar as well as the proliferation of endothelial cells from fetal bovine heart (FBHE) and human umbilical veins (HUVEC). Finally, we also tested whether SW-13 cells transiently transfected with pPTNwt would secrete autocrine-acting growth factor by plating them into soft agar. As shown in Fig. 2C, a 4-fold higher colony number in soft agar was observed from the transient pPTNwt versus pPTNmu-transfected SW-13 cells.
Obviously, the activity of the transiently secreted growth factor is sufficient to induce an autocrine growth stimulation in this assay. Taken together, these data indicate that biologically active PTN is secreted into the media of human cells after transient pPTNwt transfection.

PTN Expression and Biological Activity after Stable Transfection of Human Cells
In a separate set of experiments we studied the effects of stable PTN transfection to rule out potential artifacts of a After a 24-h incubation, labeled media were loaded onto two separate heparin-Sepharose columns, unbound material was washed off, and bound proteins were then eluted with 2 M NaCI. The eluted proteins were analyzed by a 15% SDS-PAGE (for details see "Materials and Methods"). The autoradiogram was obtained after 3 days of exposure. The arrolvhead indicates the only detectable band that differs between lane 1 and 2. B, immunoassay for P T N of conditioned media from mutant and wild-type transfected 293 cells after heparin affinity chromatography as in A. A rabbit polyclonal antiserum against the NHZ-terminal sequence of P T N was used. The control represents the absorbance reading after incubation without any antigen (for further details see "Materials and Methods"). C, assay for growth stimulatory activities of PTN. Conditioned media were collected from the transiently pPTNwt-and pPTNmu-transfected 293 cells and loaded onto heparin-Sepharose columns as in A. The bound proteins were eluted by 2 M NaCl and then assayed for soft agar colony formation of SW-13 cells and normal rat kidney fibroblasts (NRK) as well as for mitogenic activity on fetal bovine heart ( F B H E ) and human umbilical vein endothelial (HUVEC) cells. Spontaneous colony formation of transiently transfected SW-13 cells in soft agar was assayed in addition (for details see "Materials and Methods"). transient overexpression. SW-13 cells were chosen for these studies, since they do not form colonies in soft agar unless stimulated by added FGFs or P T N (3). Our earlier studies have also shown that expression of secreted K-FGF in these cells is sufficient to make them clonogenic in vitro and tumorigenic in vivo (19). From the transient transfection stud-Pleiotrophin Function and Expression ies, we selected pPTNwt and pPTNmu stable transfectant SW-13 cell lines using (3-418, and we obtained several cell lines of clonal origin in addition to the mass-transfected SW-13 population. Northern Blot-From our construct we expected a 0.8-kb P T N transcript in the pPTN-transfected cells which is clearly distinct from the constitutively expressed P T N mRNA of 1.4 kb in untransfected PTN-positive cells. A Northern blot with RNA from the PTN-positive MDA-MB-231 breast cancer cells indeed shows this latter transcript (Fig. 3A, lane 1 ) and no transcript in the parent SW-13 cells (lane 2). A high level of a 0.8-kb transcript hybridizing with the PTN probe was observed in the pPTNwt-and pPTNmu-transfected SW-13 cells (lanes 3 and 4 ) , and no signal was seen from SW-13 cells transfected with the vector alone without a P T N insert ( l a n e 5 ) . Obviously the transfected SW-13 cells produce a high steady-state level of mRNA with both the mutant and wildtype PTN constructs, and the transfection itself does not induce transcription of the endogenous gene.
Growth in Soft Agar-Stably transfected SW-13 cells were placed in soft agar and Fig. 3B illustrates that cells harboring pPTNwt grow large colonies relative to the pPTNmu cells. Stable expression of P T N obviously leads to a phenotype of SW-13 cells that can grow anchorage-independently due to a n autocrine action of the growth factor.
Heparin affinity chromatography of conditioned media from the stable pPTNwt and pPTNmu-transfected SW-13 cells was carried out to compare the affinity profile with that of native PTN originally purified from supernatants of MDA-MB-231 cells (3). We used the same step gradient of 0.4, 0.9, and 2 M NaCl employed previously to elute growth factor activity from the column. As can be seen from Fig. 3C, pPTNwt-transfected SW-13 cells release a heparin-binding growth factor into their media that is eluted by 0.9 M NaCl. This heparin affinity profile is identical to that of the original P T N purified from MDA-MB-231 cells. It is clearly distinct from acidic FGF and basic FGF which require higher salt concentrations for their elution (3).
Clonal Cell Lines-G-418 resistant clonal SW-13 cell lines transfected with pPTNwt or pPTNmu were generated and expanded to assay for P T N gene expression, biological activity of the secreted product, and for tumorigenicity in athymic nude mice. Fig. 4 shows a dose-response curve of conditioned media from three different clonal cell lines that express similar high levels of PTNwt or PTNmu mRNA. As is apparent, only the pPTNwt expressing cells secrete soft agar growthstimulating activity for the parent SW-13 cells into their media. These experiments with stable transfections lend additional support to the growth factor activity of P T N shown above in the transient transfection studies.

Tumorigenicity of S W-13 Cells Transfected with PTN
Our earlier studies showed that expression of an autocrineacting growth factor can make the nontumorigenic SW-13 cells grow into tumors in athymic nude mice (19,21). Since SW-13 cells respond to PTN by forming colonies in soft agar (see above), we used PTN-transfected SW-13 cell lines to assess the potential role of this growth factor for tumor development in vivo. Different clonal SW-13 cell lines transfected with pPTNwt and pPTNmu were injected subcutaneously into athymic nude mice. We chose clonal cell lines with different levels of P T N expression to test whether tumorigenicity is dependent on the level of P T N gene expression. P T N mRNA steady-state levels of the clonal cell lines were assessed by RNase protection assays. Four clones expressing PTNwt at different levels were chosen as positive controls, whereas a high pPTNmu expressor clonal cell line served as a negative control (see Table I). Tumor formation was checked on a weekly basis.
After 8 weeks no tumors were observed in the animals injected with the high pPTNmu or with the low PTNwt expressing cells (Table I) stably transfected clonal SW-13 cell lines were tested for their soft agar colony-stimulating activity in parent SW-13 cells. Clone 1 mutant and clone 8 wild type were also used in tumorigenicity studies (see Table I).

TABLE I Tumorigenicity studies in athymic nude mice wing clonal SW-13 cell lines transfected with pPTNwt and pFTNmu
Five animals per group were injected with 2 X lo6 cellsfsite. Tumors were measured 8 weeks after the injection of cells. The tumor size was calculated from the product of perpendicular diameters of the tumors (for further details see "Materials and Methods"). Obviously, if PTN is expressed at sufficiently high levels in SW-13 cells, they gain the ability to grow colonies in soft agar and also become tumorigenic in athymic nude mice.
Expression of PTN in Insect Cells Using the Baculovirus

System
To obtain large quantities of PTN, we decided to overexpress the gene in the eukaryotic Sf9 insect cells. We decided against a prokaryotic expression system, since proteins have t o be extracted under harsh conditions and in our hands P T N completely lost activity after treatment with organic solvents like acetonitrile or with the reducing agent dithiothreitol (see Ref. 3). To achieve production of biologically active disulfide bridge-containing proteins in prokaryotic systems, denaturation and refolding conditions have to be worked out empirically (see Ref. 22), and we wished to circumvent these obstacles by using the eukaryotic baculovirus/insect cell system.
We infected Sf9 insect cells with recombinant baculovirus containing a transcription unit for P T N driven by the polyhedrin promoter (for details, see "Materials and Methods"). We harvested media from these infected cells after 2-3 days and partially purified PTN from the media by heparin affinity chromatography. The predominant proteins present after elution from the column were an 18-and a 14-kDa species visualized by Coomassie staining after SDS-PAGE (not shown). We obtained the NHderminal sequence of the first 11 amino acids of both species and found that they were identical to the NH2 terminus of the secreted form of PTN (NHz-GKKEKPEKKVK) (3). Obviously, the protein gets processed properly for secretion in this system. However, this purified preparation showed no growth stimulatory activity on SW-13 cells or on NRK cells in a range between 1 and 1,000 ng/ml. We suspect that posttranslational modifications, improper processing, or folding of P T N in the insect cell system led to this biologically inactive preparation (see also ''Discussion"). It is worth pointing out that extremely high P T N concentrations of about 1 pg/ml from insect cells (16) and from a bacterial expression system (17) showed some neurite outgrowth and minor mitogenic activity, This activity could be due to traces of active P T N present in these preparations from nonmammalian expression systems. PTN mRNA in Tumor Cell Lines and Tumor Tissues PTN Expression in Cell Lines-Approximately one of four human tumor cell lines surveyed in our studies expressed the P T N gene as assessed by RNase protection assays ( Fig. 5 and Table 11). Some correlative evidence suggests that PTN gene expression could play a role in invasive and/or metastatic behavior of the cell lines or of the original tumors they were obtained from. One example is the highly tumorigenic and invasive breast cancer clonal cell line T47Dco which does express PTN, whereas the less malignant T47D parent cell line does not. Another example are the two cell lines derived from metastatic melanoma which did express PTN, whereas cell lines derived from melanocytes and from primary melanoma did not express PTN. On the other hand, expression of the PTN gene did not appear to be related to hormonal sensitivity of breast cancer cell lines, although the loss of hormone sensitivity is usually associated with a more malignant phenotype of breast cancer (23). Clinical studies will yield more conclusive evidence as to whether PTN expression in tumors is associated with a more malignant progression of the respective disease.
PTN Expression in Tumors-To address the issue of PTN expression and malignant progression, we have started to screen primary human breast cancer samples for PTN expression using RNase protection assays of freshly frozen tumor tissues. We had access to tumor samples from 27 randomly selected breast cancer patients. Seven of the samples are shown in Fig. 5 (lunes 4-10). As can be seen from the loading control (36B4 mRNA), the amount of RNA used in the different reactions was equivalent. In this experiment, PTN mRNA was detected in three of the samples at a level comparable with that of the MDA-MB-231 breast cancer cell line (cf. lane l l ) , whereas the other four breast cancer samples were negative for PTN. In our total survey so far we found P T N mRNA expressed in 16 of 27 (62%) of the breast cancer samples. Obviously, P T N is only expressed in a fraction of the tumors, and it remains to be seen in a larger survey whether P T N positive tumors indicate a different prognosis of the respective patient. It is worth pointing out that normal breast tissue, which is present in these tumor samples, does not express PTN at a significant enough level to be picked up by the RNase protection assay.
Finally, we assayed for P T N gene expression in carcinogeninduced mammary carcinoma in rats (Table 11). This is a standard animal model of breast cancer and is frequently used to evaluate the efficacy of breast cancer treatments (24,25).

DISCUSSION
Our current studies focus on the biological activities of the heparin-binding growth factor PTN and its potential as a tumor growth factor. Originally PTN was purified from developing rat brain and bovine uterine tissues and preliminary NH2-terminal amino acid sequence data of PTN were reported in 1989 (9, 12). It was postulated by different laboratories (11, 14,28) that PTN is a protein that functions during neuronal growth and development, but other data indicated that PTN also acts as a mitogen for fibroblasts (11) and for endothelial cells (5). We were the first to report the purification of PTN as a growth factor from the tissue culture supernatants of a human cancer cell line (MDA-MB-231), and we also detected active PTN in supernatants from human melanoma cells (3). We therefore postulated that P T N may play a role in the promotion of tumor growth in addition to its presumed function during neuronal growth (3).
Very recently the mitogenic activities of PTN have been questioned by different laboratories (14-17), and we will discuss this point first. In our purification protocol that resulted in the detection of PTN in conditioned media from tumor cells, we used a soft agar cloning assay to select chromatographic fractions containing biologically active growth factor (3). Although the major protein in that fraction was identified as PTN by NH2-terminal sequencing, it is conceivable that other contaminating minor proteins could contribute in part or completely to the biological activity observed. It should, however, be noted that none of the known growth factors for SW-13 cells (19) were detected by us in the purified PTN preparation by immunoblots (3), and these considerations led us to believe that PTN was the active growth factor present in the purified material. The final proof that PTN acts as a growth factor can only stem from production of recombinant material and our current study provides this evidence. The autocrine activity of PTN is apparent from experiments with SW-13 cells transfected with PTN (Figs. 2C and 3B). The paracrine activity of PTN is demonstrated in studies in which PTN is harvested from the supernatants of transfected cells and added back onto different cell lines to study its mitogenic or soft agar colony stimulating effects (Fig. 2C, 3B, and 4). We show that transient expression of PTN in human kidney cells as well as SW-13 cells results in the secretion of immunoreactive and biologically active P T N from these cells (Fig. 2). The growth factor activity was not only assessed with SW-13 cells but also using NRK fibroblasts and two different endothelial cell lines. Furthermore, stable transfection and selection of clonal SW-13 cells gave the same result, namely that PTN functions as a secreted growth factor (Figs. 3 and 4). A final experimental proof that PTN can act as a growth factor in tumors are results from animal studies which show that clonal SW-13 cells expressing PTN form tumors in athymic nude mice (Table I). It should be emphasized that in all of these studies, our negative controls were cells expressing P T N with a frame shift point mutation leading to a truncated product (Fig. 1B). We conclude that PTN acts as a tumor growth factor and that its mitogenic activities on endothelial cells make it a candidate tumor angiogenesis factor.
These results obtained with P T N expressed in human cells are in contrast with those obtained with recombinant P T N protein from insect cells (see "Results" and Ref. 16) as well as from bacteria (17). The failure to obtain biologically active protein from bacterial expression systems may be due to improper processing of this protein in prokaryotes or due to the loss of activity during extraction under the harsh conditions required to harvest the recombinant protein (see Ref. 22). This reasoning does, however, not suffice to explain the negative data obtained with expression of P T N in insect cells. I n agreement with data published very recently from the laboratory of H. Rauvala (16), we found P T N properly secreted into the media of insect cells infected with recombinant baculovirus carrying the P T N gene (see "Results"). This recombinant material was, however, not active as a growth factor, and we conclude that this expression system is not adequate to obtain large quantities of biologically active material for further studies. We wish to emphasize that the lack of biological activity of proteins obtained from the baculovirus expression system is not uncommon and different laboratories have attributed this to poor processing of the protein (29), improper folding (30,31), or to inadequate posttranslational modifications like phosphorylation (32,33) or glycosylation (34). We are currently testing if the recombinant P T N from the baculovirus system can be activated by refolding the protein (22, 30) or by removing potential posttranslational modifications with phosphatases (35) or with glycosidases. It is worth pointing out that a t extremely high P T N concentrations of about 1 gg/ml from insect cells (16) and from bacterial expression (17) some neurite-outgrowth and minor mitogenic activity was actually observed. This activity could be due to traces of active PTN present in these preparations and makes it worthwhile trying to recover a fully active preparation. Alternatively, transfection with pPTNwt provides high levels of P T N in supernatants from human SW-13 cells (2-10 ng/ ml; see Fig. 4) and could be used as a source for recombinant protein. These levels are about 10-fold above those of conditioned media from the breast cancer cell line MDA-MB-231 originally used in our purification of P T N (3).
An aspect closely related to the above discussion of the biological activity of PTN is the potential significance of P T N gene expression for human tumor growth. Our survey of different human cell lines showed that PTN is expressed in about one fourth of the cell lines and it appears as if gene expression might be related to a more malignant phenotype (Table I1 and Fig. 5). This is true for the comparison between cell lines derived from melanocytes and from primary melanoma versus metastatic melanoma (36, 37) as well as for the comparison between T-47Dwt and the more invasive and tumorigenic T-47Dco subclone (see Table 11). Whether or not P T N expression supports a more malignant phenotype is currently being addressed by expressing the gene in PTNnegative tumor cell lines.
We have expanded our studies with cell lines to tissues from primary human tumors and found 16 of 27 breast cancer samples positive for PTN (Table 11, Fig. 5). Since breast cancer samples contain mixtures of normal and of tumor tissue, the fact that 11 of the samples were negative for P T N mRNA indicate that P T N is not present in normal breast tissue a t levels high enough to be detected with RNase protection assays. P T N detected in the other 16 breast cancer samples is thus most likely derived from the tumor cells and not from activated normal host parenchyme. In a pilot study with four lung cancer samples, we found all of them highly positive for P T N mRNA in contrast to five samples from normal lung that showed no or very faint bands in the RNase protection assay.' In line with the observations in human tumor tissues, carcinogen-induced rat mammary carcinoma showed P T N mRNA in eight of eight samples ( Table 11). The data with the carcinogen-induced rat mammary cancer samples indicate that the PTN gene is turned on during the transforming events, and it is tempting to speculate that PTN gene expression is the switch from the nonangiogenic to the angiogenic phenotype (27,38,39). Careful studies monitoring tumor angiogenesis and PTN gene expression after carcinogen exposure in this model will be required to obtain conclusive evidence. Finally, whether or not PTN gene expression in human cancer indicates a less favorable outcome of the disease will require a larger number of samples from patients with longterm clinical follow-up.

CONCLUSION
We demonstrate that PTN functions as a tumor growth factor in vivo, stimulates endothelial, fibroblast, and epithelial cells in vitro, and may contribute to a more aggressive phenotype of tumors. Further clinical studies will have to show if the PTN gene expression observed by us in a small set of human tumors could signify a more malignant phenotype.