Evidence That Fibulin Family Members Contribute to the Steroid-dependent Extravascular Sequestration of Sex Hormone-binding Globulin*

Sex hormone-binding globulin (SHBG) binds steroids in the blood but is also present in the extravascular compartments of some tissues. Mice expressing a human SHBG transgene in the liver have human SHBG in their blood. In these animals, human SHBG accumulates within the stromal matrix of the endometrium and epididymis. This is remarkable because these tissues do not express the transgene. Human SHBG administered intravenously to wild-type mice in the presence of estradiol is rapidly sequestered within the endometrial stroma, and this prompted us to search for SHBG interacting proteins. Yeast two-hybrid screens revealed that fibulin-1D and fibulin-2 interact with the amino-terminal laminin G domain of SHBG. These interactions were verified in GST-pull down assays in which human SHBG bound the carboxyl-terminal domains of fibulin-1D and fibulin-2 in a steroid-dependent manner, with estradiol being the most effective ligand, and were enhanced by reducing the N-glycosylation of human SHBG. Like human SHBG, fibulin-1 and fibulin-2 concentrate within the endometrial stroma. In addition, SHBG co-immunoprecipitates with these fibulins in a proestrus uterine extract. These matrix-associated proteins may therefore sequester plasma SHBG within uterine stroma where it can control sex-steroid access to target cells. Given the interplay between fibulins and numerous proteins within the basal lamina, interactions between SHBG and matrix proteins may exert novel biological effects.

Biologically active androgens and estrogens are transported in the blood by proteins, including sex hormone-binding globulin (SHBG), 4 and SHBG concentrations are the main determinant of the plasma distribution of these sex steroids between the protein-bound and non-protein-bound fractions (1). This is important because only non-protein-bound or "free" steroids diffuse from blood vessels to their target cells, according the "free hormone hypothesis" (2). However, this process is influenced by physiological and anatomical variables, including blood flow and transit time in a given tissue, the extent of tissue vascularization and vascular permeability, the lipid composition of cell types in various compartments within tissues, and the extracellular matrix composition (2). These are important considerations because in some organs the free steroid must traverse several cell types within the stroma, such as lipid-rich adipose cells, as well as the complex macromolecular structure of the basal lamina, before gaining access to target epithelial cells.
Occupancy of the SHBG steroid-binding site varies depending on physiological state and circulating steroid levels, with almost all sites being unoccupied in children prior to puberty, and substantially greater occupancy of the sites in men than in women (3). High resolution crystal structure analyses have shown that each subunit of the human SHBG homodimer contains a high affinity binding site for both testosterone and estradiol within its amino-terminal laminin G-like (LG-4) domain (4 -6) and that androgens and estrogens are oriented differently within this site (7). Moreover, occupancy of the steroid-binding site by different types of steroids influences the conformation of the LG-4 domain (6), especially in relation to a flexible loop structure that covers the steroid-binding pocket (8). Interestingly, this region of the LG-4 domain of SHBG represents either a ligand-binding site or a macromolecular interaction domain in several structurally related proteins, such as neurexin, serum amyloid P component, and laminin-2␣ (9).
Steroid ligand-dependent interactions between plasma SHBG and cell membranes have been reported (10 -12). The current consensus is that unliganded SHBG interacts with a receptor located within the plasma membranes of some steroid-sensitive cell types, including breast (13) and prostate cells (14), and that occupancy of the SHBG steroidbinding site propagates an intracellular signaling event (15). However, there is also evidence that SHBG interacts with human endometrial plasma membranes in a steroid ligand-specific manner (10). Although the identity of plasma membrane "receptors" for SHBG remains obscure, ligands (e.g. protein S and gas-6) for members of the tyrofamily of orphan tyrosine kinases receptors comprise a carboxyl-terminal region, which resembles the tandem LG-4/LG-5 domain structure of SHBG and functions as their receptor-binding site (16). Recently, it has been reported that mouse megalin binds and internalizes SHBGsteroid complexes via an endocytotic mechanism (17), as observed earlier in human breast cancer cells (18) and epithelial cells of the rat epididymis (19).
During our studies of mice that express human SHBG transgenes (20), we have noted that human SHBG accumulates in the stroma of several sex steroid-sensitive tissues in which the transgenes are not expressed, such as the endometrium and epididymis, and that the sequestration of human SHBG from the blood by the endometrial stroma of mice is enhanced specifically by estrogen. These observations support an active mechanism to enhance sex steroid availability to specific target tissue compartments. To explore the molecular mechanisms underlying this extravascular sequestration of plasma SHBG, we set out to identify proteins that interact with human SHBG in yeast two-hybrid screens of a human prostate cDNA library, by using the LG-4 domain of SHBG as "bait," and we then screened candidate proteins for their ability to interact with human SHBG in vitro. These experiments revealed that the matrix-associated proteins, fibulin-1D and fibulin-2, interact specifically with SHBG in a sex steroid-dependent manner, and co-localization of these proteins in the endometrial stroma suggests that fibulin family members contribute to the ligand-dependent sequestration of human SHBG within the extravascular compartments of some sex steroid-sensitive tissues in vivo.

EXPERIMENTAL PROCEDURES
Animals-Mice expressing human SHBG transgenes and their wild-type (C57BL/6 x DBA2) littermates have been characterized previously (20). For some studies, female mice were ovariectomized using a standard operating procedure. Mice were maintained under standard conditions with food and water provided ad libitum, and the experimental procedures were approved by the Animal Use Subcommittee of the University Council on Animal Care at the University of Western Ontario.
Immunohistochemistry-After transcardial perfusion with phosphate-buffered saline, tissues of transgenic or wild-type mice were fixed in situ by perfusion with 4% paraformaldehyde followed by immersion in the same fixative overnight at 4°C. Fixed tissues were embedded in paraffin wax for routine histology and immunohistochemistry, as described previously (20). The primary antibodies used for immunohistochemistry included rabbit anti-human SHBG antibodies purified by immunoaffinity chromatography (21), rabbit anti-human fibulin-1, which also recognizes mouse fibulin-1 isoforms (sc-20818, Santa Cruz Biotechnology Inc., Santa Cruz, CA) and fibulin-2 antisera (generously provided by R. Timpl). Immunoreactive proteins were located using the EnVision TM ϩ System, HRP (DAB or AEC) detection system (DakoCytomation California Inc., Carpinteria, CA).
In Vivo Experiments-Wild-type BALB/c female mice were treated by gavage with 2.5 g of estradiol/g of body weight in 100 l of sesame oil daily for 3 days. This pretreatment was done to circumvent cycle-dependent changes in uterine morphology and to ensure that the uteri were at the same condition of estrogen stimulation. On day 4, mice received either 100 l of sesame oil vehicle alone, or 2.5 g of estradiol or testosterone/g of body weight in the same volume of sesame oil by gavage. Two hours later, anesthetized mice were injected via the tail vein with either 100 l of physiological saline containing 130 g of unliganded human SHBG, or the same amount of human SHBG in the presence of a 10-fold molar excess of either estradiol or testosterone. Two hours postinjection, blood samples were taken for a saturation steroid-ligand binding capacity assay of SHBG concentrations (22), and the mice were sacrificed for immunohistochemistry, as described above.
Yeast Two-hybrid Experiments-Coding sequences from the aminoterminal (LG-4) domains of human (residues 1-205) and rabbit (residues 1-199) SHBG (23,24) were amplified by PCR using specific oligonucleotide primer pairs for insertion into the pAS1 vector at BamHI and SalI sites in-frame with the DNA-binding domain of GAL4. These GAL4-fusions were used as baits in independent screening assays of a human prostate cDNA library fused to the GAL4 activation domain in pACT2 (Matchmaker cDNA library HL4037AH, BD Biosciences, Mississauga, Ontario, Canada). Saccharomyces cerevisiae strain Y190 was co-transformed with the pAS1 bait constructs and the prostate cDNA library. Selection for the HIS3 reporter gene was performed on plates lacking tryptophan, leucine, and histidine and containing 25 mM 3-amino-1,2,4-triazole. After 1 week, colonies were restreaked and tested for ␤-galactosidase activity by using a 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-gal) colony-lift filter assay (as recommended in the Matchmaker manual). Plasmids from positive colonies were extracted and amplified in Escherichia coli JM109 to obtain the DNA sequences, which were then used to screen public databases with BLAST 2.0. In this initial screen, the following proteins were identified as interacting with both human and rabbit SHBG baits: filamin A, translation initiation factor F, metallothionein-II (MT-II), protein KIAA116, and fibulin-2. A separate screen using only the human SHBG bait confirmed potential interactions with filamin, MT-II, and fibulin-2 and added the carboxyl-terminal domain of fibulin-1D as a potential interacting partner.
The specificity of interaction between human fibulin-1D and fibulin-2 and SHBG was evaluated using other baits. For this experiment, coding sequences for the LG-4 domains of human, mouse, and rabbit SHBG, the LG-5 domain of human SHBG (residues 205-373) and the amino-terminal LG domain of protein S were amplified by PCR and cloned into pAS1 for co-transformation into the yeast Y190 with either the human fibulin-1D or fibulin-2 pACT2 clones identified in the initial yeast two-hybrid screen. The interaction assay was monitored by both HIS3 selection and LacZ reporter gene activation, as described above.
To map the region of human fibulin-2 that interacts with SHBG, deletion mutants of fibulin-2 were prepared by PCR. Amplified sequences were inserted between the EcoRI and XhoI sites of pACT2 and co-transformed into the Y190 yeast strain together with a pAS1 vector containing coding sequence of the human SHBG LG-4 domain, and two-hybrid interactions were monitored by HIS3 selection.
GST Pull-down Assays-The cDNAs encoding residues 495-677 of fibulin-1D and 1063-1157 of fibulin-2 were excised from the pACT2 clones obtained from the human prostate cDNA library by EcoRI and XhoI digestion. These cDNAs were then inserted into the same sites of a pGEX-KGK expression vector (Amersham Biosciences) for transformation into E. coli BL21 and the production and purification of GST fusion proteins, as recommended by Amersham Biosciences. Purified GST-fusion protein (25 g) or the same amount of GST was immobilized on glutathione-Sepharose, and incubated with diluted transgenic mouse serum containing ϳ20 nM human SHBG in the presence or absence of 100 nM sex steroid (estradiol or 5␣-dihydrotestosterone) or with a culture medium from Chinese hamster ovary cells that contained a similar amount of recombinant human SHBG or human SHBG mutants lacking specific N-linked carbohydrate chains in the presence of 100 nM estradiol (25). Samples were incubated in 20 mM Tris-HCl, pH 8.0, containing 2.5 mM CaCl 2 , 0.02% Nonidet P-40, and 0.2 mg/ml bovine serum albumin overnight at 4°C. Sepharose-bound complexes were sedimented by centrifugation, and the beads were washed three times with the same buffer (without bovine serum albumin). Proteins bound to the beads were extracted by boiling in SDS-loading buffer, resolved by SDS-PAGE, and transferred by electrophoresis onto a nitrocellulose membrane for immunochemical detection of human SHBG, as described previously (25).
Co-immunoprecipitation Assays-Soluble protein was obtained from uteri of human SHBG transgenic mice at diestrus or proestus. Tissues were first minced and then homogenized by sonication in extraction buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 10% glycerol, 2 nM estradiol) containing complete EDTA-free protease inhibitor mixture (Roche Diagnostics). The protein concentration of extracts was measured and 400-g aliquots were incubated with antibodies specific to the fibulin-1 or fibulin-2 (0.5 g/reaction) at 4°C overnight. Protein A/G Plus-agarose beads (Santa Cruz Biotechnology) were then added, and the reactions were further rocked at 4°C for 2 h to allow antibodyprotein complexes to be captured by the agarose beads. Immunoprecipitated protein complexes were recovered by centrifugation, supernatants containing the unbound proteins were removed, and the beads were washed with ice-cold extraction buffer without protease inhibitor. After three washes, the agarose beads were resuspended in SDS sample buffer, and the immunoprecipitated protein complexes were recovered and resolved by 10% SDS-PAGE, followed by Western blot analysis for the target proteins.
RNA Analyses-Reverse transcriptase (RT)-PCR experiments were performed from 2.5 g of total RNA extracts of tissues taken from adult male and female transgenic mice carrying an 11-kb human SHBG transgene. Human SHBG and mouse fibulin-1D and fibulin-2 cDNAs were first prepared by reverse-transcription and were then amplified in a PCR with primers specific for the corresponding nucleotide sequences (Table 1). Mouse ribosomal RNA S16 was also amplified from each sample as a control for RNA integrity and for the RT-PCR.
Statistical Analysis-Data are reported as mean Ϯ S.D. of at least three independent experiments for all measurements. Differences between mean values were evaluated by Graph Pad Software (GraphPad Inc., San Diego, CA) by using the unpaired Student's t test.

RESULTS
Human SHBG Accumulates within the Stroma of the Endometrium and Epididymis of Transgenic Mice-Immunohistochemical analyses of tissues from several lines of transgenic mice that express human SHBG transgenes (20) revealed pronounced accumulations of immunoreactive human SHBG in the stroma of the endometrium and epididymis (Fig. 1A). This observation was remarkable because northern blotting experiments had indicated that these tissues lack human SHBG transcripts (not shown). Under high power magnification, the human SHBG in the endometrial and epididymal stroma appears to be associated with matrix fibers (Fig. 1A, iii and v, respectively). In light of a report that megalin is concentrated on the luminal surface of mouse endometrial epithelial cells (17), it is also of interest to note there was absolutely no human SHBG in this location.
Sequestration of Human SHBG from Blood by the Mouse Endometrial Stroma Is Steroid-Ligand-dependent-Because human SHBG transcripts are undetectable in the uteri of our transgenic mice, we performed an experiment to determine whether the human SHBG within the endometrial stroma originates from the blood. To accomplish this, we administered human SHBG into the tail veins of estrogen-primed, wild-type female mice in the presence or absence of testosterone or estradiol. Two hours later the uteri were removed and prepared for immunohistological analyses. The serum concentrations of human SHBG in these animals at the time of sacrifice were ϳ600 nM, which is similar to that in the blood of mice that express human SHBG trans-genes (26) and is about twice as high as in the blood of pregnant women (1) or women taking oral contraceptives (27).
As expected, human SHBG was undetectable in the uteri of vehicletreated control mice. Although low levels of human SHBG are present in the uteri of animals injected with the purified protein in the absence steroid ligands or the presence of testosterone, pronounce accumulations of human SHBG were consistently observed in mice in which the protein was administered in the presence of estradiol (Fig. 1B). Moreover, the human SHBG accumulated primarily in the stromal matrix and was concentrated toward the basal lamina of the endometrial epithelium in the animals treated with SHBG in combination with estradiol (Fig. 1B).
The Carboxyl-terminal Domains of Fibulin-1D and Fibulin-2 Interact with the Amino-terminal LG-4 Domain of SHBG in a Yeast Twohybrid Assay-To minimize the false positives encountered in yeast two-hybrid assays, we screened the yeast expression library independ-

Primer name Sequence
Human SHBG forward 5Ј-GGGAGTGATTTTTTATGG Human SHBG reverse 5Ј-TGGACATACTCAGCTTCA Mouse fibulin-1D forward 5Ј-TGCCGACGAGGTTACCAG Mouse fibulin-1D reverse 5Ј-GGCGCACGACACCCACAG Mouse fibulin-2 forward 5Ј-CGAAGGGCAATGAGGAGGGCTACT Mouse fibulin-2 reverse 5Ј-AAGGAATGGAATGTCAGCTGTAGC Mouse S16 forward 5Ј-TCCGCTGCAGTCCGTTCAAGTCTT Mouse S16 reverse 5Ј-GCCAAACTTCTTGGATTCGCAGCG FIGURE 1. Localization of immunoreactive human SHBG in mice expressing human SHBG transgenes versus wild-type controls (top panels) and non-transgenic BALB/c mice injected with human SHBG (bottom panels). A, uterus and epididymis sections from wild-type (i and iv) and transgenic mouse (ii, iii, and v) carrying the human SHBG transgene were analyzed by immunohistochemistry using a rabbit antiserum specific to human SHBG (i-ii at 100ϫ magnification, iii-v at 400ϫ magnification). Immunoreactive human SHBG appears brown. B, uteri from Balb/C mice injected with either purified human SHBG alone (i.v. SHBG), or the same amount of human SHBG saturated with either estradiol (i.v. SHBG ϩ E2), or testosterone (i.v. SHBG ϩ T ), or a saline control (no i.v. SHBG), were subjected to immunohistochemistry to localize human SHBG within various tissue compartments (brown).
ently with two bait constructs that comprised the LG-4 domains of human or rabbit SHBG. Although several potential SHBG-interacting proteins were consistently identified in a series of yeast two-hybrid screens (see "Experimental Procedures"), our observation that human SHBG is sequestered within the stromal matrix of the endometrium and epididymis focused our attention on the matrix-associated proteins, fibulin-1D and fibulin-2, which were both identified as potential binders of the LG-4 domain of SHBG (Fig. 2). In both cases, only the carboxyl-terminal regions of fibulin-1D and fibulin-2 were present in the pACT2 clones identified from the human prostate cDNA "prey" library. To confirm the specificity of these interactions, we examined the ability of these fibulin-1D and fibulin-2 sequences to interact with the LG-4 domains of SHBG from several species, as well as several other LG domain structures (i.e. the corresponding LG-4 domain within human protein S and the carboxyl-terminal LG-5 domain of human SHBG) in yeast two-hybrid assays. The results show that these fibulin-1D and fibulin-2 sequences interact only with the LG-4 domains of human, mouse, and rabbit SHBGs (Fig. 3).
The region of fibulin-2 identified as interacting with SHBG is much smaller than the fibulin-1D sequence and spans only 95 residues within the carboxyl-terminal fibulin-type module (Fig. 2). Moreover, when amino-and carboxyl-terminal deletions of this fibulin-2 sequence were tested in a yeast two-hybrid assay, we identified a minimal interaction domain spanning residues 1081-1157 within the fibulin-2 sequence (Fig. 4).
Interactions between the Carboxyl-terminal Domains of Fibulin-1D and Fibulin-2 and Human SHBG Are Steroid-Ligand-dependent-A GST pull-down assay was used to verify the interaction between the carboxyl-terminal domains of fibulin-1D or fibulin-2 and human SHBG and to determine the effects of steroid ligands on these interactions. Instead of using pure SHBG for these studies, we used diluted transgenic mouse serum containing ϳ20 nM human SHBG as a means of testing the specificity of the interactions in a semiphysiological context, i.e. in the presence of other plasma proteins. The results show the carboxylterminal domains of these fibulin family members interact selectively with human SHBG (Fig. 5A) and that this is more efficient in the presence of steroid-ligand, with estradiol promoting a stronger interaction than the androgen 5␣-dihydrotestosterone (Fig. 5B).
Loss of N-linked Oligosaccharides from Human SHBG Enhances Its Ability to Interact with Fibulin-1D and Fibulin-2-Because plasma SHBG exists as mixture of isoforms characterized primarily by the differential utilization of specific N-linked glycosylation sites (25), we tested the ability of SHBG mutants lacking specific N-linked glycosylation sites to interact with the carboxyl-terminal regions of fibulin-1D or fibulin-2 fused to GST in pull-down assays. This experiment indicated that disruption of either one of the two consensus sites for N-linked glycosylation within human SHBG promoted its ability to interact with fibulin-1D and fibulin-2 sequences in the presence of estradiol, and disruption of both sites further enhanced these interactions (p Ͻ 0.005) as compared with normally glycosylated human SHBG (Fig. 6).
Uterus and Epididymis of Human SHBG Transgenic Mice Lack Human SHBG mRNA but Contain Fibulin-1D and Fibulin-2 mRNA-A semiquantitative RT-PCR analysis was used to verify the absence of human SHBG mRNA in the uterus and epididymis from mice expressing an   11-kb human SHBG transgene and to assess the expression profiles of endogenous fibulin-1D and fibulin-2 in the uterus and epididymis, as compared with other tissues. As expected from our previous studies (20), human SHBG mRNA is confined to the livers and kidneys of both male and female transgenic mice, with only very low levels in the testes (Fig. 7). In addition, the results confirm a lack of human SHBG transgene expression in the epididymis and uterus, and add credence to our assumption that the presence of immunoreactive human SHBG in these tissues is because of the sequestration of SHBG from the blood. By contrast, mouse fibulin-1D and fibulin-2 transcripts are present in all the male and female tissues studied but appear to be most abundant in the uterus and epididymis (Fig. 7).

Human SHBG and Mouse Fibulin-1 and Fibulin-2 Co-localize within the Endometrial Stroma of Human SHBG Transgenic Mice, and Human SHBG Accumulates at This Site during Proestrus When Plasma Estradiol Levels Are
Highest-To determine whether the biochemical interactions observed between SHBG and fibulin-1D and fibulin-2 might be relevant in vivo, we compared the tissue distribution of human SHBG and mouse fibulin-1 and fibulin-2 in serial histology sections of a human SHBG transgenic mouse uterus. This immunohistochemical study demonstrated that human SHBG (brown), mouse fibulin-1 (pink), and fibulin-2 (pink) are confined to the endometrial stroma of the uterus, with little if any immunoreactivity in the myometrial cell types or the luminal epithelium of the endometrium. Within the endometrial FIGURE 6. N-glycosylation of SHBG influence its ability to interact with fibulin-1D and fibulin-2. A,GST-fused human fibulin-1D or GST-fused fibulin-2 were immobilized on glutathione beads and incubated with wild-type (W T ) human SHBG, or SHBG mutants (Gln-351, Gln-367, and Gln-351ϩGln-367) lacking specific N-linked glycosylation site(s) in the presence of excess estradiol. Nonspecifically bound proteins were washed off, and the SHBG (wild type or mutants) associated with the GST-fusion proteins was recovered by boiling and detected by Western blot. B, the graph shows the amount of the immunoreactive human SHBG quantified by densitometry of the corresponding Western blot from the fibulin-1D and fibulin-2 pull-down assays (mean Ϯ S.D. from three independent experiments). Significant differences from the control experiment in which wild-type SHBG was used are as indicated. *, p Ͻ 0.05; **, p Ͻ 0.005; ns, no significant difference. FIGURE 7. Semiquantitative RT-PCR analysis of human SHBG, mouse fibulin-1D, and fibulin-2 mRNA extracted from mouse uterus and epididymis. The RT-PCRs were performed using RNA isolated from various tissues of adult male and female human SHBG transgenic mice using primers specific for human SHBG, mouse fibulin-1D, and fibulin-2 mRNA sequences (see Table 1). Mouse ribosomal RNA S16 was also amplified from each sample as a control.

FIGURE 5. Steroid-dependent interactions between human SHBG and fibulin-1D and fibulin-2 in vitro.
A, GST-fibulin-1D 481-703 or GST-fibulin-2 1063-1157 immobilized on glutathione beads were incubated with charcoal-treated (stripped), diluted transgenic mouse serum containing human SHBG in the presence or absence of steroid ligands, estradiol (E2), or 5␣-dihydrotestosterone (DHT ), as indicated. Nonspecifically bound protein was washed off, and the human SHBG associated with the GST fusion proteins was recovered by boiling in SDS sample buffer and detected by Western blotting. B, the graph shows the amount of the immunoreactive human SHBG quantified by densitometry of the corresponding Western blot from the fibulin-1D and fibulin-2 pull-down assays (mean Ϯ S.D. from three independent experiments), and significant differences from the controls in which no steroid was added during the incubation are as indicated. *, p Ͻ 0.05; **, p Ͻ 0.005. stroma, staining of all three proteins was frequently concentrated toward the luminal epithelium (Fig. 8, A-C ), and immunoreactive human SHBG and mouse fibulin-1 appear to concentrate in the proximity of the basal lamina of the luminal epithelium under high power magnification (Fig. 8, D and E), whereas immunoreactive mouse fibulin-2 was distributed more uniformly throughout the endometrial stroma (Fig. 8F ).
To explore whether the accumulation of human SHBG in the endometrial stroma may be relevant under physiological conditions, female mice expressing a human SHBG transgene were sacrificed at specific stages of the estrus cycle (diestrus, proestrus, estrus, and metestrus), as defined by cytological examination of vaginal smears (28). The uteri from three animals at each stage of the estrus cycle were fixed by immersion in 4% paraformaldehyde and prepared for immunohistochemistry with an antiserum specific for human SHBG, as described above. The results were consistent in all three experiments and clearly showed that human SHBG is most abundant in the endometrial stroma during proestrus (Fig. 9B) at the time when plasma estradiol levels are at their highest (29). We also studied the levels and tissue distribution of immunoreactive fibulin-1D and fibulin-2 in these sections. Both proteins were  present in the endometrial stroma at all stages of the estrus cycle, but their abundance and distribution within the stoma varied depending on cycle stage (not shown). The distribution of human SHBG and fibulin-1D and fibulin-2 was also examined in the uterus of an ovariectomized human SHBG transgenic mouse. Under this condition of complete ovarian steroid withdrawal, the uterine stroma was substantially depleted of matrix and fibulin immunoreactivity (not shown), and human SHBG was confined to the blood vessels (Fig. 9E).
To demonstrate that SHBG interacts directly with fibulin-1 and fibulin-2 within the endometrial stroma, we performed a series of co-immunoprecipitation assays using uterine extracts from human SHBG transgenic mice taken during proestrus, when SHBG was most abundant in the stromal matrix, and diestrus, when SHBG was confined to blood vessels. When extracts were immunoprecipitated with anti-human SHBG antibodies we were unable to detect the co-immunoprecipitation of fibulin-1 or fibulin-2 (data not shown). This was not unexpected because much of the SHBG in these extracts originates from the blood vessels (as evidenced by the significant amounts of human SHBG in a diestrus uterine extract; see input in Fig. 10A), and any fibulin-bound SHBG would therefore be under-represented in the immunoprecipitate. Although substantial amounts of immunoreactive human SHBG were present in the diestrus uterine extract and presumably originated from the blood vessels, it contained only trace amounts of extractable fibulin-2, and there was no co-immunoprecipitation of SHBG (Fig. 10A). By con-trast, when antibodies against fibulin-1 or fibulin-2 were used to immunoprecipitate these proteins from the proestrus extract (Fig.  10B), fully glycosylated (larger apparent molecular size and more abundant) and partially glycosylated (smaller apparent molecular size and less abundant) human SHBG isoforms (30) co-immunoprecipitated with both of these fibulins.

DISCUSSION
The question of how lipophilic hormones, such as steroids, migrate from blood vessels to specific target cells within tissues has never been satisfactorily addressed. The free hormone hypothesis may adequately describe how plasma proteins regulate the amounts of steroids that passively diffuse into cells in intimate contact with blood, such as vascular endothelial cells or hepatocytes, but it fails to account for the complex path that steroids must take from blood capillaries to target cells in tissues, such as the breast, prostate, and endometrium.
There have been reports of SHBG in the extravascular compartments of sex steroid-sensitive tissues (31)(32)(33). Human SHBG transcripts have been detected in these tissues by sensitive PCR-based methods (34,35) but have never been shown to comprise the entire coding sequence for the SHBG precursor polypeptide. This is important because human SHBG transcripts are frequently alternatively spliced, and this results in a loss of coding sequences for the signal polypeptide required for secretion of SHBG or frame-shifts that cause premature translation termination (34). When coupled with problems inherent in ensuring the specificity of antiserum against SHBG for immunohistochemistry on human tissues, these issues have made studies of the extravascular localization of SHBG in human tissues difficult to verify or interpret.
To circumvent these problems we have used a mouse model in which the expression of human SHBG transgenes has been well characterized (20,26) and in which marked accumulations of human SHBG occur within the stroma of two tissues (epididymis and endometrium) in which the transgene is not expressed. In these mice, their endogenous shbg gene is not expressed in the liver after birth, and the only SHBG in their blood is derived from the human SHBG transgene (26). An additional and important advantage of this model is that parallel studies of wild-type mice allow us to demonstrate unequivocally the specificity of the antiserum used to localize human SHBG in the transgenic mouse tissues.
The significant accumulations of human SHBG in the epididymal and endometrial stroma of our transgenic mice suggested that it must originate from the blood. We set out to evaluate this in vivo, and we used wild-type mice for this experiment because they lack SHBG in the blood, and their tissues are negative when analyzed by immunohistochemistry using our anti-human SHBG antibodies. We chose to study the uterus because the well defined histomorphology within myometrial and endometrial compartments provided a clear indication of the celltype specificity of SHBG interactions within an extravascular context. It was also important that the animals were pretreated with estradiol for several days to preclude any estrous cycle-dependent differences in uterine morphology. The immediate effects of this type of estrogen treatment are to increase uterine blood flow, vascular permeability, and water imbibition, but this is followed by an increase in cell proliferation and growth, which was apparent within the endometrial compartments of all animals studied. In this context, it is important to note that no differences were observed in the size or morphology of the uteri taken from animals treated with estrogen on the day of human SHBG administration, as compared with the uteri from animals that received either no steroid or testosterone immediately prior to human SHBG administration. Furthermore, unlike the estrogen-enhanced accumulation of FIGURE 10. Human SHBG co-immunoprecipitates with mouse fibulin-1 and fibulin-2 in protein extracts of a proestrus uterus. Soluble protein extracts (400 g) from the uteri of human SHBG transgenic mice at diestrus (A) or proestrus (B) were subjected to immunoprecipitation (IP) assays with antibodies specific to fibulin-1 or fibulin-2. Mouse fibulin-1, mouse fibulin-2, and human SHBG in the protein complexes recovered by immunoprecipitation were detected by Western blot analysis. About 40 g of the soluble protein extracts (10% input) were also analyzed to evaluate the relative abundance of the target proteins.
human SHBG within the endometrial stroma, the different treatments had no influence on the localization of another plasma steroid-binding protein (corticosteroid-binding globulin) within the uterus, which showed a more uniform pattern of distribution with no obvious preference for the endometrial stroma (not shown). Thus, our data allow us to conclude that SHBG moves rapidly (within 2 h) and specifically from blood to the endometrial stroma where it binds to a component of the matrix. Furthermore, SHBG appears to penetrate the stromal bed most effectively when co-administered with estradiol.
The yeast two-hybrid screening strategy we designed to identify SHBG interacting proteins took into consideration our recent findings that the LG-4 domain of human SHBG contains the steroid-binding site (36). As reported by others (37), this type of screen identifies numerous potential SHBG interacting proteins (38), but the challenge is to verify these interactions in vitro and show that they are physiologically meaningful. Because SHBG appeared to accumulate within the stromal matrix of the epididymis and endometrium, we focused our attention on a limited number of extracellular protein candidates as potential interacting partners, and among these the carboxyl-terminal regions of fibulin-1D and fibulin-2 consistently provided the most robust positive yeast two-hybrid interactions. Moreover, the fact that yeast two-hybrid interactions occurred between the fibulin family members and the LG-4 domains of SHBG from several species, but not with the corresponding LG domain of protein S or the LG-5 domain of SHBG, provided us with a first indication that they were highly specific. This conclusion was reinforced by our observation that these interactions were not only steroid-ligand-dependent in an in vitro biochemical assay but were enhanced preferentially by estradiol. Taken together these data provided a strong indication that these macromolecular interactions are biologically meaningful.
Our data clearly imply that the fibulin-binding site on SHBG is located within its LG-4 domain and probably involves regions that are configured differently in the presence of androgens versus estrogens, such as the flexible loop located over the steroid-binding site (7,8). Because the carboxyl-terminal regions of fibulin-1D and fibulin-2 both failed to recognize the SHBG LG-5 domain in yeast two-hybrid experiments, we were surprised that the elimination of N-linked glycosylation sites within the LG-5 domain of SHBG significantly enhanced its ability to interact with the carboxyl-terminal regions of fibulin-1D and fibulin-2 in the GST pull-down assays. Although it is possible that disruption of N-glycosylation sites exposes hydrophobic patches on the surface of SHBG, which could promote more effective interactions with the carboxyl-terminal domains of fibulin-1D and fibulin-2, it is also possible that N-glycosylation of SHBG somehow hinders the fibulin-binding sites within the LG-4 domain of SHBG. If the latter is correct, it could be of interest because naturally occurring SHBG glycoforms lacking specific N-linked oligosaccharides (30, 39) may therefore accumulate preferentially within target tissues through an enhanced interaction with matrix-associated proteins in vivo.
As members of the fibulin family of extracellular matrix-associated proteins, fibulin-1D and fibulin-2 share a similar molecular architecture (40). They contain binding sites for numerous matrix-associated proteins and basement membrane proteins and participate as molecular scaffolds within diverse supramolecular structures (41). As in other mammalian species, alternative RNA splicing results in several isoforms of human fibulin-1, of which fibulin-1C and fibulin-1D have extended carboxyl-terminal domains (domain III) that share limited sequence similarity with domain III of fibulin-2 (41). Because our data indicate that the ligand-dependent interaction between SHBG and fibulin-2 requires only a sequence of 77 residues within domain III, it is possible that conserved residues within this region of these fibulin isoforms are critical for making contact with SHBG.
At least three other fibulin isoforms (fibulin-3, fibulin-4, and fibulin-5) have extended domain III sequences with limited sequence similarity within the region of fibulin-2 that we have identified as the SHBGbinding site (41). Although their presence within the endometrium remains to be determined, several reports have indicated that they are located primarily in the vicinity of endothelial cells (41). If these fibulin isoforms also bind SHBG, they might also participate in the initial sequestration of SHBG from the blood circulation and account for the accumulations of SHBG around blood vessels in the endometrial stroma (not shown).
The matrix-associated proteins and fibulin-1D or fibulin-2 were obvious candidates as SHBG interacting proteins for several reasons. They are both highly expressed in these tissues, they accumulate within the extracellular matrix (41), and they co-localize with human SHBG within the endometrial stroma of our transgenic mice. Interactions between fibulin-1D or fibulin-2 and SHBG within the endometrium were also of particular interest because they are both up-regulated by progesterone in human and rat endometrial stromal cells (42)(43)(44), and fibulin-1D mRNA abundance fluctuates in the human endometrium in a cycle-dependent manner (44). It was therefore interesting to note these fibulin family members co-localized with human SHBG within the endometrial stroma of our transgenic mouse model and that this was most apparent with respect to the presence of immunoreactive fibulin-1 and SHBG at the basal lamina of the luminal epithelial cells. This observation suggests to us that specific fibulin family members may not only contribute to the steroid-dependent accumulation of SHBG in the endometrial stroma but may also affect the partitioning of SHBG within different regions of the stroma. If this is correct, fibulin/SHBG interactions could provide a dynamic shuttle mechanism for the targeted transfer of sex-steroids from the blood vasculature across the stroma to luminal epithelial cells.
Although we have used a mouse model to examine how human SHBG might enter tissues such as the uterus, our observation that human SHBG accumulates at very high levels in the endometrial stroma of these mice during proestrus, suggests that this may be physiologically relevant and that detailed clinical studies may be warranted. It is also important to note that this occurs at a time when plasma estradiol levels are at their highest (29) and that at this stage of the estrus cycle there is an abundance of stromal matrix that is rich in fibulins 1 and 2. An immunoprecipitation assay of protein extracts from a human SHBG transgenic mouse uterus taken at proestrus, with anti-fibulin-1 or antifibulin-2 antibodies, resulted in the co-immunoprecipitation of human SHBG. Interestingly, the immunoreactive human SHBG in the immunoprecipitate is enriched in glycoforms (smaller apparent molecular size) lacking one N-linked carbohydrate chain (30), and this supports the notion that the preferential binding of SHBG glycosylation-deficient variants to fibulin-1D or fibulin-2 might enhance their extravascular sequestration by these matrix-associated proteins. We therefore conclude that estradiol enhances the movement of SHBG from the blood circulation and/or its accumulation in the endometrial stroma, and we propose that this is promoted by the preferential binding of specific SHBG glycoforms in complex with estradiol to fibulin family members.
The basal lamina represents an important interface between the stroma and epithelium where signals are received and transmitted between these tissue compartments, and an accumulation of SHBG at this site could have consequences extending far beyond that of simply regulating the access of sex steroids to their target epithelial cells. In this context, matrix-associated proteins provide a molecular scaffold for sig-naling molecules, such as the integrins, and form an extracellular-intracellular "bridge" for the transmission of signals between cell types separated by the basal lamina (45)(46)(47). There is also a considerable degree of cross-talk between the actions of estrogens and various integrin and growth factor signaling pathways in the extracellular matrix (48,49). Therefore, in addition to controlling the access of steroids to specific cell types within complex tissues, ligand-dependent interactions between SHBG and matrix-associated proteins, such as fibulin-1 and fibulin-2, could modulate their binding to various signaling molecules, including for instance the ␣IIb␤3 integrin receptor (50), either directly or indirectly via other key components of the basal lamina.
In conclusion, our data imply that steroid-ligand-dependent interactions between SHBG and at least two fibulin family members could contribute to the extravascular accumulation and regional distribution of SHBG within the stroma of a sex steroid-sensitive tissue. This concept clearly extends the function of plasma SHBG from one of simply transporting and regulating the plasma distribution of its sex steroid ligands and places it in a new arena in which it could be playing a more versatile role in controlling the access of steroids to their target cells in specific tissues. In this context, the ligand dependence of these interactions, coupled with dynamic changes in the expression of SHBG-interacting proteins within different tissues during different physiological states, provide for additional layers of complexity that could influence the movement of sex steroids within tissue compartments and alter their access to specific target cells. In addition, given the complex interplay between various matrix-associated proteins, steroid-ligand-dependent interactions between SHBG and various fibulin family members might initiate signals that could be transferred across the extracellular matrix and represent a novel mechanism of steroid hormone action.