Stromal inhibition of prostatic epithelial cell proliferation not mediated by transforming growth factor beta.

The paracrine influence of prostatic stroma on the proliferation of prostatic epithelial cells was investigated. Stromal cells from the human prostate have previously been shown to inhibit anchorage-dependent as well as anchorage-independent growth of the prostatic tumour epithelial cell lines PC-3 and LNCaP. Antiproliferative activity, mediated by a diffusible factor in the stromal cell conditioned medium, was found to be produced specifically by prostatic stromal cells. In the present study the characteristics of this factor were examined. It is demonstrated that prostate stroma-derived inhibiting factor is an acid- and heat-labile, dithiothreitol-sensitive protein. Although some similarities with type beta transforming growth factor (TGF-beta)-like inhibitors are apparent, evidence is presented that the factor is not identical to TGF-beta or to the TGF-beta-like factors activin and inhibin. Absence of TGF-beta activity was shown by the lack of inhibitory response of the TGF-beta-sensitive mink lung cell line CCL-64 to prostate stromal cell conditioned medium and to concentrated, partially purified preparations of the inhibitor. Furthermore, neutralising antibodies against TGF-beta 1 or TGF-beta 2 did not cause a decline in the level of PC-3 growth inhibition caused by partially purified inhibitor. Using Northern blot analyses, we excluded the involvement of inhibin or activin. It is concluded that the prostate stroma-derived factor may be a novel growth inhibitor different from any of the currently described inhibiting factors. ImagesFigure 5

Control of cellular proliferation in the prostate involves a complex interaction of different cell types with soluble peptide growth factors, (steroid) hormones and constituents of the extracellular matrix. It is likely that the array of peptide factors which play a role in the regulation of cell proliferation and differentiation affects these processes through both positive and negative control mechanisms. The direct importance of embryonic messenchyme as a mediator of the androgen-induced prostatic ductal morphogenesis, epithelial growth, secretory cytodifferentiation and function has been demonstrated convincingly (Cunha et al., 1987). Subsequently, it was shown that these interactions may have retained their integral role in the adult prostate (Cunha et al., 1987;Chung et al., 1991a). It is an intriguing observation that, upon maturation, accessory sexual organs reach a typical size and weight which is characteristic for that particular organ (Cunha et al., 1987). Likewise, administration of testosterone to androgen-deprived rats restores the size of the prostate to its normal precastration level without inducing overgrowth, even after prolonged administration (Cunha et al., 1987). At that stage, epithelial cell proliferation is low and in balance with cell death (Isaacs, 1984). Several investigators have been searching for the molecular basis of this homeostatic constraint mechanism that curtails further increase in cell number once the gland has reached its predetermined size. So far, TGF-P is the only well-kiown epithelial cell growth inhibitor that has been identified in the prostate (Barnard et al., 1990;Wilding, 1991). Its presence in prostatic tissue has been demonstrated by Northern blot analyses (Kyprianou and Isaacs, 1989;Mori et al., 1990), and elevated levels of TGF-P have recently been associated with prostate cancer (Thompson et al., 1992). In vitro studies have implicated TGF-,B as a potent inhibitor of prostatic epithelial cells, both normal and malignant (McKeehan and Adams, 1988;Wil-ding et al., 1989;Goldstein et al., 1991). As these cells have been shown to secrete TGF-P themselves (Ikeda et al., 1987;Danielpour et al., 1989), an autocrine mode of action is suggested. In vivo, TGF-P has been shown to act mainly on the rate of prostatic glandular cell death, while its expression appeared to be under negative androgenic regulation (Martikainen et al., 1990). Since these determinations were performed on whole tissue homogenates, it is still not clear whether TGF-P is produced by the stroma or the epithelium.
A recent immunohistochemical study (Truong et al., 1993) demonstrated immunoreactivity to TGF-P1I antibodies in epithelial and mesenchymal cells, but unfortunately this kind of experiment does not reveal the site of synthesis of the proteins. However, the observation that rat and human prostatic epithelial cells grown in vitro can proliferate in serumfree medium without the presence of androgens (McKeehan et al., 1984;Peehl and Stamey, 1986;Nishi et al., 1988), while androgen ablation in vivo induces a marked regression of the glandular epithelium (Isaacs, 1984), indicates a major role for the stroma in the negative control of epithelial cell proliferation. A number of investigators have studied stromalepithelial interactions in the prostate in vitro, but their reports mainly focused on the mitogenic influence of the stroma (Sherwood et al., 1988;Kabalin et al., 1989;Djakiew et al., 1991;Gleave et al., 1992;Sherwood et al., 1992). No further reports from other groups have been published on stromally derived epithelial cell growth inhibitors in the adult prostate. Remarkably, all studies demonstrating stromal stimulation of prostatic epithelial cell growth were performed under serum-free conditions. However, Kirk et al. (1981) showed an inhibition of PC-3 cell growth by lung fibroblasts, using a serum-dependent soft-agar assay. Using a similar assay, we previously observed inhibition of the clonal growth of prostatic carcinoma cell lines PC-3 (hormone independent) and LNCaP (hormone responsive) by co-cultured prostatic stromal cells (Kooistra et al., 1991). Rowley and Tindall (1987) reported growth inhibition of a bladder carcinoma cell line by medium conditioned by urogenital sinus explants in the presence of 5% fetal calf serum (FCS) and 5% (synthetic) Nu-serum. Later, it was shown that antiproliferative activity was produced by a fibroblastoid cell strain derived from urogenital sinus mesenchyme (Rowley, 1992a). Although cells were shown to express urogenital sinus-derived growth inhibitory factor (UGIF) activity for up to 5 -7 days in unsupplemented Dulbecco's Modified Eagle Medium (DMEM; changed daily), cultures failed to survive under these non-physiological conditions. The presence of the UGIF was tested in a serum-containing assay. Using physicochemical, biological and immunological methods UGIF was demonstrated to differ from known growth inhibiting factors, including TGF-P. Using the same bioassay, however, no activity could be found in a chemically defined medium developed for long-term culture that was conditioned in exactly the same manner. Only addition of steroid hormone caused the fibroblastoid cells to produce inhibiting activity that ultimately was identified as being caused by activated TGF-P and not by UGIF (Rowley, 1992b). Although a serum-free environment is ideal for testing as well as identification of regulatory peptides, certain information may be lost, as suggested by the findings reported by Kirk et al. (1981) and Rowley (1992b), and the results presented here. The prostatic stromal cell-derived antiproliferative activity we found was mediated by a diffusible factor present in serumcontaining conditioned medium (CM) of stromal cells from neoplastic lesions of (adult) prostates (Konig et al., 1987;Kooistra et al., 1991Kooistra et al., , 1995a. In the present report we show that this factor is different from TGF-P, its related peptides inhibin and activin, and from other inhibitors previously described. Awaiting further identification, we provisionally and operationally refer to this putative inhibiting factor as 'Prostate-derived epithelium inhibiting factor' (p-EIF).

Materials and methods
Stromal cell cultures Surgically obtained tissue specimens from histologically proven benign prostatic hyperplasia (BPH) and prostatic carcinoma (PC) were cut into pieces (approximately 1 mm x 2 mm), and placed in 35 mm Petri dishes (Nunc) containing 1.5 ml of basal medium: Earle's minimum essential medium (Gibco Europe, Breda, The Netherlands) supplemented with 10% FCS (Biological Industries, Beth Haemak, Israel), 2 mM glutamine, penicillin and streptomycin (all from Gibco Europe). Cultures were maintained in a humidified incubator at 37°C in 5% carbon dioxide/air. Medium was replaced twice a week. The initial halo of epithelial cells grown from these explant cultures became overgrown with fibroblast-like cells within several weeks. Subsequently, cells were detached by trypsinisation (0.05% trypsin in 0.02% EDTA) (Gibco Europe) and split in a 1:3 ratio every 2-3 weeks. In order to minimise the number of epithelial cells in our cultures we used only prostatic stromal cells of passage number 4-9 in this study. Using the monoclonal antibody NCLSD3 (Organon, Oss, The Netherlands) reacting with keratin 8, 18 and 19, we found that the number of positive cells in these cultures (BPH as well as PC) never exceeded 5%, indicating that nearly all cells were of nonepithelial origin (Kooistra et al., 1995b).

Cell lines
The prostatic carcinoma cell line PC-3, obtained from Flow Laboratories (Irvine, UK) and maintained in basal medium, was used between passage numbers 35 and 45. LNCaP-FGC cells (used at passage number 65-70), originally made available to us by Dr J Horoszewicz (Buffalo, USA), were cultured in RPMI-1640 (Gibco Europe) supplemented with 10% FCS, glutamine and antibiotics. The mink lung cell line MLCCL 64 (Holley et al., 1983) was obtained from the NIH (Frederick, MO, USA) and cultured in DMEM-high glucose (Gibco Europe) supplemented with 7.5% FCS and antibiotics. All cultures were shown to be free of mycoplasma contamination by staining with Bisbenzimide (Hoechst dye 33258) obtained from Sigma (St Louis, MO, USA).
Concentration and partial purification CM of different passages was pooled and fractionated by ammonium sulphate precipitation. This was performed through a stepwise increase in the level of saturation by adding solid ammonium sulphate (Sigma) to the medium at 0°C under continuous stirring (Dixon, 1953). At every 10% rise the solution was centrifuged (10000g, 4°C, 20min) and the pellet dissolved in PBS (Gibco Europe). Samples from supernatant and pellet were extensively dialysed (Spectrapor 3, cut-off 3500 dalton; Spectrum Medical, Los Angeles, CA, USA) at 4°C against PBS and MEM successively, and stored at -20°C after sterilisation through a 0.451 m membrane (Schleicher & Schuell, 's-Hertogenbosch, The Netherlands). Osmolarity was checked to be 275-325 mosmol, using a Roebling osmometer (Vogel, Giessen, Germany). Using this 'first-step' purification procedure, 75% of contaminating protein could be discarded (Kooistra et al., 1995a).
Physicochemical characterisation Concentrated and partially purified CM was diluted with PBS to a protein concentration of 2 mg ml-'. Trypsin sensitivity was tested by incubation with trypsin (T 8003, Sigma) at a concentration of 100 ltg ml-' for 2 h at 37°C, followed by the addition of soybean trypsin inhibitor (STI) (T 9003, Sigma) at a final concentration of 200 ILg ml-'. As a control, equal amounts of trypsin and STI were preincubated in PBS for 30 min at 20°C and subsequently incubated with the samples for an additional 2 h at 37°C. The reducing agent, dithiothreitol (DTT) (Sigma) was added to each sample at a final concentration of 130 mM. Incubation was performed on a rocker platform for 60 min at 20°C. Control samples were incubated without DTT. Acid treatment was performed by acidifying samples to pH 1.5 with 2 M hydrochloric acid for 1 h at 20°C, then neutralising by 2 M sodium hydroxide. Control samples were adjusted to the same volume as test samples, using 2 M sodium chloride. Heat sensitivity was tested in the presence of 100 mM Hepes, at 56°C for 30 min, or at 100°C for 3 min. Samples were incubated in closed glass tubes in a rocking water bath. Controls for all tests mentioned above were obtained by similar treatment of basal medium (diluted to a protein concentration of 2 mg ml-'). After incubation, all samples were subsequently dialysed at 4°C against PBS and MEM, and sterilised by filtration over a 0.45 jsm pore membrane. For determination of growth inhibition (MTT test), 180 jil of sample and 20 ,ul of FCS were added at day 0. Stability at a range of temperatures was tested by heating aliquots of CM for 5 min at indicated temperatures. In order to keep the pH at the same level throughout the incubation, Hepes was added to a concentration of 100 mM. As a control, basal medium and serum-free basal medium were treated likewise. Samples were subsequently dialysed against PBS and MEM and the inhibition of cell growth was tested after sterile filtration, using the MTT test or the [3H]thymidine incorporation assay.
Neutralisation tests TGF-i1 from human platelets and TGF-P1 neutralising polyclonal antibodies raised in rabbits were prepared at the Hubrecht laboratory, Utrecht, The Netherlands (Van den Eijnden-Van Raaij et al., 1990a). Porcine platelet TGF-P2 was obtained from Sandoz (Basle, Switzerland), while TGF-P2-specific (neutralising) antibodies, also from rabbit, were purchased from R&D Systems (Minneapolis, MN, USA). Concentrated, semipurified inhibitor was diluted to 1 mg ml-', preincubated with antibody (30 Lg ml-') for 1 h at 37°C and subsequently added to wells containing PC-3 cells (MTT test). Controls were obtained by preincubation with non-immune serum IgG. Activity of immune serum IgG was tested by incubation of TGF-P11 (10 ng ml-' in MEM, 0.1% bovine serum albumin) with equal amounts of antibody.
Tests with TGF-P2 antibodies were performed with 80 lg ml-' antibody, and for controls 20 ng ml' TGF-P2 was used. Optimal amounts of antibody (saturable activity) were determined by incubating inhibitor with different amounts of antibody.
Protein determination Protein concentrations were measured with the Bio-Rad Protein Assay Kit (Bio-Rad Labs, Veenendaal, The Netherlands). A 1:1 mixture of the albumin and globulin solutions was used as a standard.

Growth inhibition assays
MTT test Inhibition of cell growth was determined by means of a colorimetric assay based on the reduction of a tetrazolium salt, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT), to a coloured formazan product (maximum absorption at 560 nm) by mitochondrial enzymes present only in living, metabolically active cells (Romijn et al., 1988). Cells were harvested by trypsinisation, resuspended in fresh culture medium, and plated in a volume of 0.1 ml per well in 96-well microtitre plates (Costar). Inoculum: PC-3, 2000 per well; LNCaP, 5000 per well; MLCCL-64, 1500 per well. The next day (day 0), 0.1 ml of sample was added, giving a 50% dilution of the sample. At day 3, 30 lA of a 5 mg ml-' solution of MTT (Sigma) in PBS was added to each well. After a 4 h incubation at 37'C in 5% carbon dioxide/air, the medium was carefully sucked off and the purple dye was dissolved in 0.1 ml of dimethylsulphoxide (DMSO) (Merck, Darmstadt, Germany). Plates were placed on a plate shaker for 5 min, and the absorbance at 560 nm was read using a Flow Titertek Multiskan plate reader. Unless otherwise stated, eight replicate wells were used for each sample. Wells containing medium but no cells served as blanks. Results are expressed as percentage of maximal growth from day 0, obtained in fresh basal medium. To determine the number of cells at this point, one extra plate was measured at day 0.
[3H]Thymidine incorporation This was measured using 5000 MLCCL-64 cells per well in 24-well plates (Nunc). The mink lung carcinoma cell line CCL-64 is known to be very sensitive to TGF-ps (Tucker et al., 1984;Danielpour et al., 1989). Cells were plated on day 1 in 1 ml of DMEM (Gibco Europe) supplemented as described. After 4 h, 0.1 ml of sample was added. At day 3, 0.5 ,Ci of [3H]thymidine (Amersham, UK) in 0.1 ml of Ham's F12 medium was added to each well and plates were incubated for another 16 h. Monolayers were washed four times with PBS, then fixed in methanol for 15 min at room temperature and dried in air. Cells were then dissolved in 1 ml of 1 M sodium hydroxide (30 min, 37°C), transferred to a scintillation vial and radioactivity was counted.
Northern blot analyses After removal of culture medium, stromal cells were frozen in solid carbon dioxide/ethanol, and stored at -80°C. Total RNA was isolated using an acid guanidinium thiocyanate-phenol-chloroform extraction procedure (Chomczynski and Sacchi, 1987). Of each sample, 40 gtg of total RNA was denatured in formamide/formaldehyde at 55°C for 15 min before electrophoresis on denaturing 1% agarose/ Stromal inhibiton of prostaic epithelium A Kooistra et al 429 formaldehyde gels. After electrophoresis, RNA was blotted on Hybond N+ (Amersham, UK) by diffusion. Filters were baked for 2 h at 80°C, and subsequently prehybridised for 2 h at 42°C in a hybridisation solution containing 50% formamide, 9% (w/v) dextran sulphate, 10 x Denhardt's [1 x -Denhardt's contains 0.02% (w/v) Ficoll, 0.02% (w/v) polyvinylpyrrolidone, 0.02% (w/v) bovine serum albumin], 5 x SSC (1 x SSC contains 0.15 M sodium chloride, 0.015 M sodium citrate, pH 7.0), 10 mM sodium phosphate (pH 6.8), and lOO1lgmlm' denatured salmon sperm DNA. Probes for hybridisation were labelled with 32P by random oligonucleotide labelling (Feinberg and Vogelstein 1983), denatured in boiling water for 5 min and added directly to the hybridisation solution. Following 48 h of hybridisation at 42°C, filters were washed to a final stringency of 0.1 x SSC, 0.1% SDS, 50°C. Filters were exposed to Amersham Hyperfilm-MP (Amersham, UK) at -70°C for various lengths of time, using an intensifying screen. Probes used were described by Esch et al. (1987) and Derynck et al. (1985). As a control for equal amounts of RNA, a hamster actin cDNA probe was used.

Statistical analysis
The statistical significance of differences between individual treatment groups was calculated using Student's t-test. Differences were considered statistically significant if the twotailed P-value was smaller than 0.05.

Results
Effect ofprostatic stromal cell CM and partially purified inhibitor on the growth ofprostatic tumour epithelial cells We previously reported the inhibition of anchorageindependent growth of the hormone-insensitive prostatic carcinoma cell line PC-3, as well as the hormone-responsive prostatic carcinoma cell line LNCaP, by co-cultured prostatic stromal cells (Kooistra et al., 1991). Figure 1 shows that CM from prostatic stromal cells was capable of inhibiting the growth of PC-3 cells in monolayer cultures up to approximately 50%. Growth in PBS was significantly better than that found in dialysed CM, indicating that the inhibition was not due to depletion of nutrients or of growth-promoting factors. Concentration and partial purification of CM resulted in a potent solution reaching inhibition levels up to 80-90% at protein concentrations less than 5 mg ml'. In a comparative experiment, bovine serum albumin was shown to interfere with cell growth only at higher concentrations, demonstrating that the observed inhibition of semi-purified inhibitor was not merely due to a high protein content. As a control, basal medium and CM from skin fibroblasts were E s Time (days) Figure 1 Inhibition of PC-3 cell growth by prostatic stromal cell CM. The y-axis gives the absorbance at 560 nm as determined by the MTT test (see Materials and methods). All samples were dialysed against basal medium without FCS. 0, Basal medium; 0, PBS; A, conditioned medium. Error bars = s.d.

Stromal inhibiton of prostabic epithelium
A Kooistra et al 430 fractionated in a similar way. No inhibiting activity was found in these preparations. As can be seen in the dilution curves in Figure 2, LNCaP cells are almost equally sensitive to the antiproliferative activity as PC-3 cells. Preliminary studies on primary cultures of BPH and prostatic carcinoma epithelium showed these cells to be responsive too.

Biochemical characterisation
The inhibitory activity of partially purified p-EIF preparations was tested under different conditions in an attempt to elucidate the nature of the putative inhibitor. The fact that the inhibitor could be precipitated by ammonium sulphate already suggested that a protein is involved. This idea was supported by the observation that trypsin digestion resulted in a statistically significant reduction in inhibitory activity (Figure 3a). Treatment with DTT resulted in a complete disappearance of activity (Figure 3b requirement of S-S bridges for biological activity. Acid treatment of partially purified inhibitor significantly reduced the inhibiting activity ( Figure 3c). Surprisingly, treatment of basal medium resulted in the generation of inhibiting activity, which may explain the observation that inactivation of the stromal cell-derived inhibitor did not restore growth to 100%, although inactivation of growth-promoting substances might also have contributed to this effect. Figure 3d shows that treatment of the concentrated fraction of inhibitor at 56°C for 30 min resulted in a small, but statistically significant, decrease in antiproliferative activity. This observation suggested p-EIF to be heat labile. However, heating at 100°C restored the inhibiting activity to levels found before treatment, while similar effects were seen after heating basal medium. In order to look at this aspect more carefully, we performed similar experiments on CM, using a wider range of temperatures. The increasing levels of inhibition generated in basal (serum-containing) medium by treatment at higher temperatures are clearly demonstrated (Figure 4a). However, a tendency of the stromal cell inhibitor to lose its inhibiting capacity (observed after heating at 56°C and 70C) was noted.

Northern blot analysis
To explore the possibility that TGF-P or TGF-13-like subs-  Figure 5). However, since stromal cell CM from Wilms' tumour and normal kidney tissue did not inhibit PC-3 cell growth (Kooistra et al., 1995a), this does not necessarily imply that sufficient amounts of (active) TGF-P are secreted into the medium. We also examined the involvement of activin and inhibin in cultured prostatic stromal cells. As shown in Figure 5, no P-A/B chain or a-chain mRNAs were detected in prostate derived stromal cell cultures, indicating that neither activin nor inhibin was synthesised in these cells. Hybridisation occurred only with the probe for the P-A chain and was restricted to cultures derived from normal kidney tissue. The presence of P-A mRNA in human kidney fibroblasts has not been reported before (Meunier et al., 1988).
Bioassay using MLCCL-64 cells for detection of TGF-P activity To test further the production of active TGF-P by prostatederived stromal cell cultures, we used a bioassay with the TGF-p-sensitive mink lung carcinoma cell line MLCCL-64.
For comparison, the inhibition of PC-3 cell growth by TGF-P and partially purified inhibitor was also determined. Both TGF-PI and TGF-P2 were found to inhibit PC-3 cell growth ( Figure 6a). The concentrated solution of inhibitor had to be diluted ten times to cause an inhibition similar to that of 5 ng ml-' TGF-,B. As might be expected, TGF-P1I and TGF-132 had a stronger antiproliferative effect on MLCCL-64 cells than on PC-3 cells (Figure 6b). However, no significant inhibition of MLCCL-64 cell growth was observed with CM or with serial dilutions of the partially purified inhibitor. These data demonstrate that CM-mediated inhibition of PC-3 cell growth was not caused by TGF-P. Those findings were confirmed by the highly sensitive [3H]thymidine incorporation assay on MLCCL-64 cells that appeared to be slightly stimulated by fresh-frozen CM as well as by partially purified p-EIF (not shown). However, after prolonged storage (more than 2 months) CM tended to gain some inhibitory activity on both cell types when compared with fresh basal medium, probably as a result of activation of latent TGF-P.
Using this [3H]thymidine incorporation assay on MLCCL-64 cells increasing amounts of active TGF-13 were also detected in samples from basal medium as well as from CM after heating (Figure 4b).
Effects of TGF-P neutralising antibodies In order to confirm the absence of active TGF-P, we performed MTT tests on PC-3 cells using neutralising antibodies against TGF-P1 and TGF-P2. As shown in Figure 7a, preincubation of TGF-P1 with polyclonal antibodies raised against purified TGF-P1 neutralised its effect completely. Those antibodies, however, did not cause a significant decline in the level of inhibition induced by partially purified p-EIF. Tests with TGF-,B2-specific antibodies gave essentially the same results (Figure 7b). These findings, again, demonstrated that the putative inhibitor present in the CM was different from TGF-P. Since PC-3 cells are known to secrete predominantly TGF-P2 in their CM (Ikeda et al., 1987), we were also interested in the influence of TGF-P2 antibodies on the growth of PC-3 cells themselves. Figure 7b shows that these antibodies did not improve the growth of the PC-3, indicating that there was no active 'autocrine' TGF-P2 present that could be recognised by the antibodies.

Discussion
Following the observation that epithelium loses its growth capacity when separated from the stroma (Franks et al., 1970), it has become increasingly clear that both autocrine and paracrine factors produced by epithelial and stromal cells play an important role in the local control of prostatic growth. It has been demonstrated that urogenital sinus mesenchyme plays a major role in the (androgen-induced) development of the gland (Cunha et al., 1987); subsequently it was shown that these interactions may have retained their integral role in the adult prostate (Cunha et al., 1987;Chung et al.,1991a,b). These observations indicated that the development of cancer and other disease states is likely to involve a loss of coordination or other alterations in such interactions.
McNeal (1978) suggested the reversion of prostatic stroma to an 'embryological state' inducing inappropriate epithelial proliferation in benign prostatic hyperplasia. Tenniswood (1987) hypothesised the existence of two growth-stimulating factors and one growth inhibitor, produced in stroma and epithelium, that control growth and differentiation. He also suggested that the development of BPH may be caused by continuously elevated expression of one or more of these factors. Indeed, several epithelial cell growth-promoting peptides have been identified in the prostate and in prostatederived epithelial cell cultures (Thompson, 1990;Story, 1991), and some of these have been positively identified as being produced (also) by stromal cells, suggesting a role in stromal-epithelial interactions (Story et al., 1989;Djakiew et al., 1991;Gleave et al., 1992). The fact, however, that prolonged androgen administration to castrated animals does not induce the gland to grow beyond its predetermined size (Sugimura et al., 1986), together with the observation that withdrawal of androgens decreases epithelial cell number only in the in vivo situation where stroma is present (Isaacs, 1984;McKeehan et al., 1984), suggests a role for stromally derived epithelial cell growth inhibitors. To date, the only well-studied epithelial cell growth inhibitor identified in the prostate is TGF-P (Wilding, 1991). The transforming growth factor P family includes a group of closely homologous proteins (Massague, 1987). Three distinct molecular forms, designated TGF-31, TGF-P2 and TGF-P3, have been identified in mammals (Barnard et al., 1990). Most cells, including stromal cells, secrete TGF-P as a latent complex which can be activated by treatment with heat, acid or proteases such as plasmin (Lawrence et al., 1985;Rowley, 1992b). Also, commercially available sera have been shown to contain latent TGF-P (Childs et al., 1982). The expression and activation appears to be influenced by steroid hormones (Kyprianou and Isaacs, 1989;Martikainen et al., 1990), as was shown also for urogenital sinus mesenchyme (Rowley, 1992b). The expression of TGF-P mRNA in human prostatic tissue (TGF-P1 and -P2; Mori et al., 1990) as well as in rat ventral prostate (TGF-p1; Kyprianou and Isaacs, 1989) has been demonstrated by Northern blot analyses. Recent immunohistochemical studies of diseased human prostates demonstrated TGF-P1 immunoreactivity in both epithelial and mesenchymal cells (Truong et al., 1993). In vitro studies with TGF-13I have shown that this factor inhibits proliferation of prostatic carcinoma cell lines PC-3, DU 145 and LNCaP (Schuurmans et al., 1988;Wilding et al., 1989;Goldstein et al., 1991), while PC-3 cells were found to secrete latent TGF-1 (predominantly 02) in the culture medium (Ikeda et al., 1987;Danielpour et al., 1989).
Two other members of the TGF-P family of growth and differentiation factors are activin and inhibin (Massague, 1987). Both factors have previously been shown to compete for TGF-P binding to pituitary tumour cells (Cheifetz et al., 1988). Furthermore, TGF-p-like properties of activin in developmental biology were recently demonstrated (Van den Eijnden-Van Raaij et al., 1990b). Both activin and inhibin can interfere with cell growth (Gonzales-Manchon and Vale, 1989;Hedger et al., 1989), and inhibin-like proteins have been demonstrated in seminal plasma and prostate tissue (Shah and Sheth, 1991).
Using a two-layer soft-agar system, we previously observed an inhibition of the clonal growth of prostatic carcinoma cell lines by prostatic stromal cells. It was shown to be mediated by a diffusible factor present in the CM (Kooistra et al., 1991(Kooistra et al., , 1995a, which we referred to as p-EIF (Konig et al., 1987). In the present work we investigated the possibility that p-EIF is identical to one of the well-known epithelial growth inhibitors mentioned above. Since TGF-,B mRNAs were expressed in our stromal cell cultures ( Figure 5), active TGF-P could have been produced (among latent TGF-,B) by the stromal cells (Rowley, 1992b). On the other hand, the active fraction of inhibitor might consist of 'fibroblast-activated' latent TGF-P, for instance, present in the serum (Childs et al., 1982;Antonelli-Orlidge et al., 1989). Our data show that p-EIF has several properties in common with TGF-P, including reversibility of inhibition (Hebert and Birnbaum, 1989; Kooistra et al., 1995a) and sensitivity to trypsin and reducing agents. Elaboration of inhibiting activity under acidic conditions ( Figure 3c) is known to occur in solutions containing latent TGF-,B (Lawrence et al., 1985). From the observations described above we may conclude that latent TGF-P was derived from the serum present during conditioning and/or secreted by the stromal cells themselves. Since activation was observed in stability tests performed on concentrated partially purified inhibitor preparations, we have to conclude that latent TGF-P co-precipitated with p-EIF at 40-50% ammonium sulphate saturation. Similar considerations apply to heat treatment (Figure 3d). However, the tendency of the inhibitor to lose its inhibiting capacity upon heating (significant at 56°C and 70°C), was probably obscured at higher temperatures by the generation of active TGF-,B from the solution (Figure 4). This observation appeared to discriminate our inhibitor from the TGF-,B family of growthmodulating factors. However, since the sample is not highly purified yet, we realise that the possibility of co-precipitation of active inhibitor adhering to other denatured products cannot be ruled out. Nevertheless, further strong biological and immunological evidence was provided that p-EIF is different from these inhibitors. In particular, the observation that the antiproliferative effect of semipurified p-EIF was not altered by P1 and P2 neutralising antibodies (Figure 7) ruled out the presence of active TGF-P1 and 132 in these preparations.
Recently, growth inhibition of a bladder carcinoma cell line as well as PC-3 cells by serum-containing medium conditioned by a fibroblastoid cell strain subcultured from urogenital sinus was reported (Rowley and Tindall, 1987;Rowley, 1992a). Physicochemical properties of the reported factor (UGIF) suggest that it is different from that found in the present study. However, it should be kept in mind that in both cases tests were performed on crude preparations limiting the value of these observations. Limonti et al. (1992) reported the antiproliferative effect of luteinising hormonereleasing hormone (LHRH) agonists on LNCaP cells. Specific binding sites were demonstrated, while receptor as well as peptide levels were thought to be negatively regulated by androgens. However, since stromal cell-derived inhibitory activity was not lost upon dialysis (cut-off 3.5 kDa), these peptides can be ruled out as candidates for the reported inhibitory activity on the basis of their molecular size. Interferons (IFNs) have also been shown to inhibit growth of prostatic epithelial cells (Deshpande et al., 1989;Goldstein et al., 1991;Okutani et al., 1991), and, among other cell types, fibroblasts have been recognised as a rich source of interferon (arbitrarily called IFN-P) (Vilcek et al., 1987). We believe, however, this is not a likely candidate either, for two reasons. First, untreated fibroblasts produce subeffective concentrations of IFN, and only treatment with 'inducers' (e.g. virus infection or double-stranded RNA chains) leads to the secretion of detectable amounts of IFN-P in the CM. Second, LNCaP cells were shown not to be inhibited by IFN-,B (Goldstein et al., 1991), while we demonstrated inhibition of anchorage-dependent and anchorage-independent growth of LNCaP cells by prostatic stromal cell CM.
In conclusion, the data presented in this paper demonstrate that adult prostate-derived stromal cells cultured from neoplastic lesions produce a unique factor, tentatively called 'prostate-derived epithelium inhibiting Factor' (p-EIF). On the basis of its spectrum of biological activity as well as its physicochemical and immunological properties, p-EIF can be discriminated from previously described growth inhibitors, including the TGF-p-related proteins inhibin and activin (Massague, 1987), and from interferon beta (Goldstein et al., 1991). The organ-specific production and the lack of inhibition on all non-prostatic epithelial cell lines tested, emphasise the importance of this inhibitor (Kooistra et al., 1995a). It would be very interesting to investigate whether this factor is the major 'brake' on the prostate, controlling epithelial cell proliferation and thus prostatic size. Several aspects concerning the production of p-EIF are still under investigation. Preliminary data suggest that serum-derived factors are required for optimal production and/or secretion. We failed to demonstrate significant effects of steroids (testosterone, dihydrotestosterone or oestradiol) in this respect however. Further knowledge of the factors influencing its production will facilitate purification and characterisation. Antibodies raised against the purified factor would provide us with an excellent tool for studying its role in normal physiology and its possible involvement in the development of neoplasia.