Sex hormone-binding globulin in the human prostate is locally synthesized and may act as an autocrine/paracrine effector.

Sex hormone-binding globulin (SHBG) is a plasma protein synthesized and secreted by the liver. Its initial description stemmed from its ability to bind estrogens and androgens and its capacity to regulate the free concentration of the steroids that bind to it. Additionally, it participates in signal transduction for certain steroid hormones at the cell membrane. It binds with high affinity to a specific membrane receptor (R(SHBG)) in prostate stromal and epithelial cells, wherein the SHBG.R(SHBG) complex forms. An appropriate steroid binds to this complex and results in increases of intracellular cAMP. These two disparate functions of SHBG, regulation of the concentration of free steroids in plasma and signal transduction in selected tissues, raise the question of how its synthesis and secretion might be regulated so as to best perform these two disparate functions. In this paper we demonstrate that SHBG is produced in human prostate cancer cell lines (LNCaP, DU 145, and PC 3) as well as in cultured human prostate epithelial and stromal cells. In addition, in tissue sections of human prostate, we demonstrate the presence of SHBG (immunocytochemistry) and SHBG mRNA (in situ hybridization). These observations are consistent with the hypothesis that SHBG, destined to participate in signaling at the cell membrane, is locally regulated and produced.

The hepatically derived (1) plasma protein, sex hormonebinding globulin (SHBG) 1 has two apparently disparate functions. It is the major regulator of free estrogens and androgens in plasma (2)(3)(4); it is an initiating component of a signaling system for estrogens and androgens at the cell membrane that functions via a G protein (5) and cAMP (6 -9). The details of the working of these systems generate a model that is in need of clarification in a number of crucial respects. In this paper we present data that deal with the source of the SHBG for signaling at the cell membrane. These data support a model in which the source of SHBG for signaling is not the plasma, but rather the target cells themselves. SHBG participates in signaling by first binding to its receptor (R SHBG ) on selected cell membranes. After this occurs, the binding to SHBG of an agonist steroid stimulates the rapid production of cAMP. There are a number of important details that constrain this event. SHBG cannot bind to its receptor if it is already bound to a steroid (whether an agonist or antagonist). It is an allosteric protein that, after it binds a steroid, assumes a conformation that prohibits binding to R SHBG (10). The sequence of events must be exact: SHBG must first bind to R SHBG forming SHBG-R SHBG and the steroid must bind to that complex. If the steroid is an antagonist (as are most steroids that bind to SHBG) then no signal is generated; if the steroid is an agonist, e.g. 3␣-androstanediol, estradiol (11)(12)(13), a signal (cAMP) ensues.
None of the foregoing explicitly addresses the source of the SHBG that binds to R SHBG . Until now, its origin has been presumed to be the liver, which synthesizes and secretes SHBG into the plasma (1). In women (in plasma) about 50% of SHBG is not bound to steroids and in men unbound SHBG constitutes about 20% (14). Although a number of factors cause substantial increases or decrease in plasma SHBG, these changes take place in days to weeks not hours to minutes (2). Furthermore, these changes are chronic ones that persist; there is little short term (hours) variation in plasma SHBG. Thus, if plasma were the source of SHBG for binding to R SHBG , the substrate for any rapidly mediated SHBG signaling would have to be via either change in steroid concentration or by rapid changes in R SHBG at the cell surface. There are some data in the literature that imply the possibility that SHBG is produced locally. Reports of SHBG immunologic activity in a variety of tissues, with one exception (15), are tempered by the presumption that it is there by virtue of having been taken up from the plasma (16 -20).
In this paper we demonstrate that SHBG is synthesized and secreted by tissues and cells that are known to have R SHBG , thus creating the potential for the local regulation of SHBG signaling by autocrine and paracrine mechanisms.

EXPERIMENTAL PROCEDURES
General-Human prostatic LNCaP, PC 3, and DU 145 cancer cell lines were obtained from the American Type Culture Collection (ATCC; Manassas, VA) and maintained in recommended growth media. Cells were refed every 3-4 days and split 1:4 to 1:8 when nearing 80% confluence. The ALVA-41 cell line was obtained from Drs. S. Loop and S. Ostensen of the American Lake Veterans Affairs Medical Center (Tacoma, WA) and maintained as described previously (21). Highly purified human SHBG was prepared and its purity verified as described previously (6).
Isolation of Stromal and Epithelial Cells from Prostatic Tissues-Prostatic tissues were obtained from patients with benign prostatic hyperplasia at the time of prostatectomy and immediately brought to the laboratory under sterile conditions. Tissues were minced, mixed with RPMI 1640, containing 1% collagenase and 0.05% DNase, and incubated for 30 min (37°C, 5% CO 2 , 95% air) with occasional shaking. * This work was supported in part by grants from the Starr Foundation and the Lowey Family Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The mixture was centrifuged at 3000 rpm for 10 min, the supernatant discarded, and the tissues redigested for 1 h. The dissociated cells were suspended in RPMI 1640, containing 10% fetal bovine serum and 1 mM sodium pyruvate, and plated in 75-cm 2 Corning flasks. The medium was changed after 6 days and every 5 days thereafter. The primary culture was enriched for stromal cells by supplementing the growth medium with 10% fetal bovine serum and 20 ng/ml insulin (12). Stromal cell identity was confirmed by staining with a monoclonal antibody specific to human vimentin intermediate filament (Novacastra, Newcastle, UK) (Fig. 1). Epithelial cell colonies were morphologically identified from duplicate cultures plated in cell culture dishes. These colonies were isolated and propagated in RPMI 1640, 5% fetal bovine serum, and 1 mM sodium pyruvate. After multiple passages, four cell lines from four different tissues were spontaneously immortalized. In their 20 th passage, each has a doubling time of about 2 days. The epithelial identity of the cells was confirmed by staining with a monoclonal antibody specific to human cytokeratins 5, 6, 8, and 18 intermediate filaments (Novacastra). Color development was as described under "Immunocytochemistry (Immunofluorescence)." Polymerase Chain Reaction (PCR)-Total RNA from each cell type was isolated using the TRIzol reagent according to the manufacturer's instructions (Invitrogen). One g of total RNA was used to prepare first strand cDNA using the Advantage RT-for-PCR kit (CLONTECH, Palo Alto, CA) according to their instructions. For RT-PCR, the PCR reaction mixture contained: 1.5 mM MgCl 2 , 0.2 mM dCTP, dGTP, dTTP, and dATP, 2.5 units of Taq polymerase (Platinum Taq polymerase; Invitrogen), 1 M concentration of each primer and the cDNA, in a total volume of 50 l of PCR reaction buffer. The PCR reaction was carried out in a GeneAmp PCR, system 9600 (PerkinElmer). PCR amplification was performed as follows: denaturation at 94°C for 30 s, annealing at 55°C for 1 min (54°C for the androgen receptor), and extension at 72°C for 2 min for 30 cycles (25 cycles for ␤-actin). The final PCR products were electrophoresed on 1.2% agarose, visualized with ethidium bromide, and digitally imaged using the Electrophoresis Documentation and Analysis System 120 (Eastman Kodak Co.).
Immunocytochemistry (Immunofluorescence)-LNCaP, PC 3, and DU 145 cells were grown on NUNC CC2 chamber slides (Nalgene Nunc International, Naperville, IL) for 3 days, rinsed in PBS, and fixed in 4% paraformaldehyde (PFA) for 30 min Ϯ 0.1% Triton X-100 (permeabilized). The cells were then rinsed with 0.1 M glycine/PBS and Trisbuffered saline (TBS) and treated with SuperBlock Blocking Buffer (Pierce) for 1 h Ϯ 0.1% Triton X-100. Following blocking, the cells were incubated overnight (4°C) with a monoclonal antibody against SHBG (10 g/ml, 5B2) in SuperBlock Blocking Buffer that was diluted 1/5 in TBS or 10 g/ml mouse IgG1 (Southern Biotechnology Associates, Birmingham, AL) that was similarly diluted. Cells were rinsed again in TBS for 5 min and then incubated (30 min, room temperature) with the green fluor, Alexa-488 coupled to rabbit anti-mouse IgG1 (Molecular Probes, Eugene, OR) that was diluted 1:500 in SuperBlock Blocking Buffer. Cells were rinsed in TBS and then twice in PBS for 5 min each, fixed for 10 min in freshly prepared 4% PFA, rinsed again as before, and then mounted with Vectastain mounting medium containing the blue nuclear dye DAPI (Vector Laboratories, Burlingame, CA). In experiments in which SHBG and/or 2-methoxyestradiol were added, LNCaP cells were grown as usual in eight-well CC2 chamber slides. After 2 days, the cells were washed with PBS. To appropriate wells we added: 0.5 ml of Opti-MEM I, 0.5 ml of OPTI-MEM containing 100 nM highly purified, stripped (devoid of steroids) SHBG, or 0.5 ml of Opti-MEM containing 1.0 M 2MeOE 2 and 100 nM SHBG. After incubating overnight in 5% C0 2 , 95% air, the chamber slide was washed three times with PBS and processed for immunofluorescence as above. Imaging was with either an Olympus BX60 microscope, attached to a MicroImager II digital camera (QIMAGING, Burnaby, Canada), or an LSM 5 Pascal confocal microscope (Zeiss Inc).
Immunocytochemistry (ABC Method)-Frozen samples of prostate were cut into 5-m sections and mounted onto glass microscope slides. Samples were fixed in 4% PFA for 30 min, blocked with normal horse serum for 1 h, and incubated overnight (4°C) with rabbit anti-SHBG (24) (1:10,000 dilution) in normal horse serum. Slides were developed by the ABC method (Vector ABC elite system) using 3,3Ј-diaminobenzidine counterstained with hematoxylin and mounted with Vectamount (Vector Laboratories) permanent mounting medium. Photographs were taken with a 35-mm camera, mounted on an Olympus BX60 microscope (ϫ400), and digitized.
In Situ Hybridization-Paraffin-embedded prostate sections were deparaffinized according to the protocol described in the Biogenex Super Sensitive In-Situ Hybridization Kit. The sections were placed in an RNase block for 5 min, fixed in 4% PFA for 30 min, prehybridized for 60 min (37°C), and then rinsed in PBS. A 521-bp SHBG-(628 -1148) biotinylated probe was prepared by PCR incorporation of biotin-14-dCTP (Invitrogen) into the SHBG PCR product. The probe was added to the slides, which were heated (95°C) for 10 min, and then incubated overnight (37°C) in a humidified chamber. Slides were developed using the ABC method (Biogenex Super Sensitive In-Situ Hybridization Kit), using 3,3Ј-diaminobenzidine as the substrate, and counterstained with hematoxylin before mounting using Vectamount (Vector Laboratories) permanent mounting medium. Photographs were taken with a 35-mm camera mounted to a BX60 microscope (Olympus) and digitized.

Generation of Primary Cultures of Human Prostate Stromal
and Epithelial Cells-We generated four human prostate stromal cell cultures (HPSC1-HPSC4) and four human epithelial (HPEC1-HPEC4) cell lines with samples of tissue taken from patients undergoing prostate surgery. Immediately after removal, samples were minced, cultured, and enriched for either stromal or epithelial cells. All human cells were characterized immunocytochemically with the stromal cell-specific vimentin antibody and the epithelial cell-specific pancytokeratin antibody (25). The stromal cultures stained with the vimentin antibody but not with pancytokeratin antibody, whereas the epithelial cells stained with the pancytokeratin antibody but not with the vimentin antibody (Fig. 1A).
As expected, the stromal cells expressed the stromal cellspecific KGF gene, but not the epithelial cell-specific gene, PSA (Fig. 1B). In addition, none of the epithelial cell lines expressed KGF. To the best of our knowledge, epithelial cell lines that constitutively express PSA have not been reported. Thus, it was surprising that even a single cell line (HPEC4) expressed this gene. Its identity was confirmed by restriction mapping with the 4-base cutters, HinfI and HaeIII. The slightly larger and less intense background fragments from HPSC1, -2, -4, and HPEC1-3 gave different restriction patterns (data not shown). Both cell types expressed the androgen receptor.
SHBG mRNA Is Expressed in Human Prostate Cancer Cell Lines and Primary Cultures-That cultured human prostate cells synthesize SHBG mRNA was shown by RT-PCR analysis of established prostate cancer cell lines, the four stromal cultures, and the four epithelial cell lines described above (Fig. 2). cDNA templates were prepared from total cellular RNA isolated from: human LNCaP, DU 145, PC 3, and ALVA 41 prostate cancer cell lines; the HPSC1-HPSC4 stromal cultures; HPEC1-HPEC4 epithelial lines, whole prostate; and Hep G2 cells, which synthesize and secrete SHBG (1,26). As an additional positive control, we examined a stromal culture transfected with an SHBG expression vector (TrHPSC3). SHBG primers were selected to amplify a 521-bp band encompassing sequences from exons 5-8 (22). A 521-bp SHBG RT-PCR transcript was generated in samples from the three human prostate cancer cell lines, the four stromal cultures (including the transfected one), the four epithelial lines, whole prostate, and the Hep G2 cells. In addition, a smaller 313-bp fragment, consist-ent with the alternatively spliced SHBG transcript (SHBG T ) lacking exon 7 (27)(28)(29), was generated at a lower apparent abundance in all but the HPEC 1 RT-PCR samples. DNA sequence analysis confirmed the identities of these two bands (data not shown). Unexpectedly, we also observed a smaller 178-bp RT-PCR band in the same samples (data not shown), which sequence analysis showed to result from direct splicing of exon 5 to exon 8. The significance of this alternatively spliced transcript is not clear at the this time. Thus, SHBG mRNA is expressed in both prostate epithelial and stromal cells, as well as in prostate cancer cell lines.
SHBG Protein Expression in LNCaP, DU 145, and PC 3 Cells-LNCaP, DU 145, and PC 3 prostate cancer cell lines, permeabilized with Triton X-100 after fixation, were probed with both monoclonal (Fig. 3) and polyclonal antibodies to SHBG. Identical results were obtained with a polyclonal antibody (64-4) generated in this laboratory (24) and seven different monoclonal antibodies, four obtained from Dr. Niver P. Sahakian (Diagnostic Products) and three from Lewis et al. (30). Monoclonal antibody 5B2 (Diagnostic Products) proved to be the best with which to work, but qualitatively similar results were seen with all the other antibodies. To rule out contamination of the growth medium, we assayed it for SHBG with an enzyme-linked immunosorbent assay (Alpha Diagnostic International, San Antonio, TX) whose limit of detection for SHBG is 0.2 nM. None was detected. (For comparison, the average SHBG concentration in the plasma of men is 25 nM (31).) SHBG is visualized clearly in the cytoplasm, but not the nucleus, of all three cell lines. Although we have chosen representative photomicrographs, we should emphasize that LNCaP cells both stain more intensely than the other lines and also (less apparent) show such staining in a greater fraction of the cells than either DU 145 or PC 3 cells. Nonpermeabilized LNCaP cells showed some membrane staining for SHBG (Fig. 5A2), consistent with secretion of SHBG and binding to R SHBG . Western blot analysis confirmed that SHBG is synthesized by LNCaP, DU145, and to a lesser extent, PC3 cells (data not shown).
SHBG Protein and mRNA Localization in Frozen and Fixed Specimens of Prostate Tissue-In normal prostate tissue,

FIG. 1. Identification of stromal and epithelial cells from human prostate.
A, microscopy (ϫ400). Primary cell cultures of stromal and epithelial cells were grown from prostate tissue obtained at surgery for benign prostatic hyperplasia. The four epithelial cell cultures (human prostate epithelial cells, HPEC1-HPEC4) were immortalized in the course of multiple passages, whereas the primary cultures of stromal cells (human prostate stromal cells, HPSC1-HPSC4) could be maintained for only 5-10 passages. To ensure the proper identification of each of the cell types, samples were stained with stromal cell-specific monoclonal antibodies to vimentin and epithelial cell-specific monoclonal antibodies to pancytokeratin. The color was developed with rabbit anti-mouse IgG1 linked to a green fluor in a medium containing the blue nuclear stain DAPI. Each of the eight samples was examined separately and gave results comparable with those shown. Furthermore, controls with nonimmune mouse IgG1 showed no staining (data not shown). Exposure of stromal cells to anti-pancytokeratin and epithelial cells to anti-vimentin revealed no discernable cross-contamination of the cell types (data not shown). B, RT-PCR. First strand cDNAs were generated from the same samples as in A and examined by RT-PCR for the expression of KGF, PSA, and the AR. KGF is a marker of prostate stromal cells (46) in culture, while PSA is specific to the prostate epithelial cell (47). The androgen receptor is expressed in both normal human prostate stromal and epithelial cells. Whole prost, whole human prostate.

FIG. 2. Expression of SHBG L and SHBG T transcripts in cul-
tures and cell lines. SHBG L and SHBG T are the two major SHBG gene transcripts, originally described in liver (SHBG L ) and testis (SHBG T ), which encode the secreted and the exon 7-lacking isoforms of SHBG, respectively. The SHBG RT-PCR primers flanked exon 7 and directed amplification of a larger 521-bp SHBG L fragment and a smaller 313-bp SHBG T fragment in the same reaction. cDNA was prepared from all of the cell lines and whole human prostate (Total prost) and examined by RT-PCR for the expression of SHBG L , SHBG T , and ␤-actin. TrHPSC3, HPSC3 transiently transfected with an SHBG L expression vector.

FIG. 3. SHBG protein expression in LNCaP, DU 145, and PC 3 cells (؋400).
These three human cancer cell lines were prepared as described under "Experimental Procedures" and exposed to a monoclonal antibody (5B2) to SHBG (B) or mouse IgG1 (A) and developed with a rabbit anti-mouse IgG1 linked to the green fluor, Alexa-488. Nuclei were stained with the blue nuclear dye, DAPI.
SHBG is intensely stained in epithelial cells, with the most intense staining seen in the luminal epithelial cells (Fig. 4B). In addition, there is staining in the stroma, but it is less intense. Because there is a substantial concentration of SHBG in plasma, the staining seen in sections of human prostate could result from simple diffusion of SHBG from plasma with subsequent cellular uptake. Indeed, that is the inference from early observations in both monkey (16) and man (17), and there is no subsequent literature that takes issue with this conjecture. To address this issue, we examined sections, adjacent to those used for SHBG expression, by in situ hybridization. In these sections of the prostate, SHBG mRNA appears in the same cells as the protein (Fig. 4, A and B). As for the SHBG protein, SHBG mRNA is most abundant in luminal epithelial cells. We also note that staining for both SHBG protein and SHBG mRNA is rather heterogeneous throughout the prostate, with areas of intense staining and other areas of zero to light staining. These observations are consistent with a model in which SHBG is synthesized and secreted primarily by epithelial cells and binds to the R SHBG on epithelial and/or stromal cells.
An Antagonist of SHBG⅐R SHBG Blocks Membrane Binding of SHBG-Using LNCaP cells, which are known to have R SHBG (6), we sought to demonstrate whether binding of SHBG to R SHBG could be demonstrated immunocytochemically (Fig. 5) and whether this binding could be inhibited by ligands for SHBG as is the case for the solubilized receptor (10). Conversely, the presence of ligands for SHBG should not alter intracellular SHBG. We stained permeabilized and nonpermeabilized LNCaP cells with monoclonal antibody 5B2. Staining in the former represents both intra-and extracellular SHBG, while staining of the latter represents SHBG outside of the cell that is inferentially bound to R SHBG . All experiments were also done Ϯ added SHBG and with both immunofluorescent and confocal microscopy. Addition of SHBG markedly increased intracellular staining in the permeabilized cells (panel B3) and membrane staining in the nonpermeabilized ones (panel A3). Most significant is the marked decrease in membrane staining (nonpermeabilized cells) seen when added SHBG is preincubated with 2-methoxyestradiol (panel A4). This steroid is known to inhibit the binding of SHBG to R SHBG (10) and to function as an antagonist to the signaling induced by the estradiol⅐SHBG⅐R SHBG complex (13,32,33). The lack of effect of 2-methoxyestradiol on SHBG staining in permeabilized cells (panel B4) supports the specificity of this effect.

DISCUSSION
That SHBG, or an SHBG-like antigen, exists in human prostate has been documented in a number of laboratories both by biochemical (34 -36) and immunocytochemical techniques (16,18). However, the uniform interpretation of these results has been that the origin of the protein was from the plasma. In contradistinction to that supposition, we show that prostate stromal cell cultures, epithelial cell lines, and prostatic tissue all express SHBG mRNA and protein. Each of the cultures and cell lines derived from patients with benign prostatic hyperplasia express both SHBG L and its exon 7-lacking isoform (SHBG T ) transcripts (Fig. 2), as do the three prostate cancer cell lines, LNCaP, PC 3, and DU 145 (Fig. 3). With the exception of HPEC 1, all of the cells express greater amounts of SHBG L than SHBG T . In the SHBG stromal cell explant transfected with an SHBG expression vector, TrHPSC3, the massive overexpression of SHBG L drowns out any detectable SHBG T ; nontransfected HPSC3 cells showed both bands (data not shown). Although not immediately apparent from the figures, we have noted that SHBG expression in tissue sections, cultured cells, and cancer cell lines is heterogeneous. Both the intensity of the staining in individual cells and the number of cells that are positive for SHBG varies with the field that is FIG. 4. Expression of SHBG protein and mRNA in fixed sections of human prostate. A, in situ hybridization (ϫ400). The section immediately adjacent to the one below (B) was prepared as in the methods and incubated with a 521-bp (628 -1148) SHBG biotinylated probe that had incorporated biotin-14-dCTP. After incubation, the slide was developed and counterstained with hematoxylin. B, immunocytochemistry (ϫ400). Frozen samples of human prostate were cut into 5-m sections, mounted onto glass microscope slides, and stained with rabbit anti-SHBG and counterstained with hematoxylin (see "Experimental Procedures" for details).

FIG. 5. 2-Methoxyestradiol suppresses membrane-associated immunofluorescence in LNCaP cells.
LNCaP cells, in serum free medium, with (permeabilized) or without (nonpermeabilized) Triton X-100, were exposed to SHBG Ϯ 2-methoxyestradiol and visualized with a monoclonal antibody (5B2) to SHBG (green) and were counterstained with DAPI (blue). The antibody control was mouse IgG. The insets are confocal micrographs that do not have a nuclear counterstain. under examination. The single paper dealing with SHBG in tissue sections of prostate cancer contained data demonstrating the same phenomenon (17).
Although we have shown that both SHBG mRNA and protein are synthesized in all the prostate material we examined, it is not conclusively clear whether the protein is translated exclusively from SHBG L . Little is known about the SHBG T isoform; its full-length sequence has not been reported, and the protein it encodes has neither been identified in vivo nor shown to be stable in vitro. SHBG L and SHBG T share a common reading frame from the middle of exon 2 through exon 6 (27,28,37) and, if stably expressed, the SHBG T protein might bind to the antibodies used in these experiments. That SHBG protein is observed on the membranes of isolated cell lines suggests that its only source is SHBG L ; the putative protein encoded by SHBG T lacks a signal peptide and should not leave the cell. However, it might serve an intracellular function. For instance, since the mRNA retains the dimerization domain (38), it could form heterodimers with SHBG L , with consequences yet to be determined.
The impact of these observations on the possible roles for SHBG is substantial. Locally regulated intracellular SHBG, in target cells for appropriate steroid hormones, could alter the intracellular free concentration of those hormones and affect their access to the classic intracellular receptors (androgen and/or estrogen). Of equal or possibly greater importance, it appears reasonable to hypothesize that the source of SHBG for membrane signaling in the prostate is that which is synthesized in the prostate itself. Thus, the regulation of its synthesis must play an important role in this signaling. This clearly is an area requiring investigation, because other than in the liver, nothing is known of it.
Heretofore, the role of SHBG as the major determinant of free androgens and estrogens in plasma has been the central force driving many clinical studies of plasma SHBG in both benign and malignant disease of the prostate (36, 39 -45). Although there is modest disagreement, alterations in plasma SHBG seem to be neither a predictor nor an accompanier of these two important diseases of the prostate. The data presented herein direct attention to targets for SHBG in tissues, not the plasma, and should form the basis for a major paradigm shift in our thinking about the biology of SHBG.