Characterization and comparison of recombinant full‐length ursine and human sex hormone‐binding globulin

Sex hormone‐binding globulin (SHBG) regulates the bioavailability of sex steroid hormones in the blood. Levels of SHBG increase markedly in brown bears (Ursus arctos) during hibernation, suggesting that a key regulatory role of this protein is to quench sex steroid bioavailability in hibernation physiology. To enable characterization of ursine SHBG and a cross species comparison, we established an insect cell‐based expression system for recombinant full‐length ursine and human SHBG. Compared with human SHBG, we observed markedly lower secretion levels of ursine SHBG, resulting in a 10‐fold difference in purified protein yield. Both human and ursine recombinant SHBG appeared as dimeric proteins in solution, with a single unfolding temperature of ~ 58 °C. The thermal stability of ursine and human SHBG increased 5.4 and 9.5 °C, respectively, in the presence of dihydrotestosterone (DHT), suggesting a difference in affinity. The dissociation constants for [3H]DHT were determined to 0.21 ± 0.04 nm for human and 1.32 ± 0.10 nm for ursine SHBG, confirming a lower affinity of ursine SHBG. A similarly reduced affinity, determined from competitive steroid binding, was observed for most steroids. Overall, we found that ursine SHBG had similar characteristics to human SHBG, specifically, being a homodimeric glycoprotein capable of binding steroids with high affinity. Therefore, ursine SHBG likely has similar biological functions to those known for human SHBG. The determined properties of ursine SHBG will contribute to elucidating its potential regulatory role in hibernation physiology.

tissues like muscle, adipose tissue, and bone [3]. According to the 'free hormone hypothesis', only free steroids can enter target cells by passive diffusion through the plasma membrane and exert their functions [4].
The plasma level of SHBG increases 45-fold in Scandinavian brown bears during hibernation [5]. During the 6-month-long hibernation period, the bear does not eat, drink, urinate, or defecate, and moves only infrequently [6]. In a period where energy conservation is critical, such an increase in SHBG suggests an unknown function of this protein during this important physiological state in bears. Contrary to humans, brown bears do not show any signs of disease from prolonged immobility and obesity [7][8][9]. High plasma SHBG levels in humans correlate with a lower risk of numerous diseases, such as metabolic syndrome, type 2 diabetes, cardiovascular diseases [10], and muscle weakness [11]-none of which are observed in the hibernating brown bear [7]. Elucidating the physiological mechanisms that protect the hibernating bear may reveal novel therapeutic targets in humans.
Human SHBG is well characterized [12][13][14][15][16][17][18]. Like the human SHBG precursor polypeptide, the ursine SHBG precursor polypeptide is 402 amino acids long, and they share 77% sequence identity. Cleavage of the human 29 amino acid secretion signal peptide results in mature human SHBG of 373 amino acids [19], but the signal peptide cleavage site and mature sequence has not been determined for ursine SHBG prior to this study. Mature human SHBG has a deglycosylated molecular weight of 40.5 kDa and forms a homodimer with a total glycosylated weight of about 90 kDa [19][20][21]. Each monomer consists of two laminin G-like domains (LG domains), joined in tandem by a short linker. The tertiary structure of the N-terminal LG domain has been solved for human SHBG, and the steroid-binding site is located close to the center of this domain in a hydrophobic core formed between two βsheets (Fig. 1A) [22]. Additionally, a yet uncharacterized receptor, located in the plasma membrane of sex steroid-responsive cells, has been reported for human SHBG [23,24]. Amino acid residues 48-57 (TWDPEG-VIFY) in the N-terminal LG domain of human SHBG were identified as the receptor-binding site (residues 51-60 in ursine SHBG) (Fig. 1B) and this region is highly conserved between species [25].
To characterize ursine SHBG, we established an insect cell-based expression system for recombinant ursine SHBG alongside the human orthologue. The expression system produced dimeric and active fulllength ursine and human SHBG. A lower steroidbinding affinity of ursine SHBG, compared to human, was observed. Except for the lower steroid affinity, ursine SHBG showed similar characteristics to human SHBG, and therefore, they likely have similar biological functions. The determined properties of ursine SHBG will contribute to elucidate SHBG's potential regulatory role in hibernation physiology.

Results and Discussion
Sequence alignment and homology modeling of ursine SHBG The ursine SHBG amino acid sequence was aligned to that of human SHBG (Fig. 2). The structure of the ursine SHBG N-terminal LG domain was predicted by homology modeling based on the structure of the human orthologue (Fig. 3). In human SHBG, steroids are anchored in the steroid-binding site through hydrogen bonds, with androgens and estrogens binding in opposite orientations. The C3 carbonyl of androgens and the C17 hydroxyl of estrogens points into the interior of SHBG and forms a hydrogen bond with the side-chain hydroxyl of Ser42 in human SHBG, which is of paramount importance for high-affinity steroid binding. Two additional hydrogen bonds are formed at the entrance of the binding pocket between the hydroxyl group at C17 of androgens or C3 of estrogens and the side chains of Asp65 and Asn82 in human SHBG, where Asp65 appears to be the most important [22,26]. These three steroid-anchoring residues are conserved in ursine SHBG (Ser45, Asp68, Asn85) (Table 1). Besides these residues, all other contacts are hydrophobic, with Phe67, Met107, and Met139 in human SHBG contributing with the largest contact surface to the steroid [26]. Phe67 is conserved in ursine SHBG (Phe70), while the two Met residues are substituted to Ile (Ile110) and Val (Val142). Trp84 in human SHBG also affects steroid-binding, although not by direct contact [26], and is substituted to Met (Met87) in ursine SHBG. SHBG forms two disulfide bridges (Cys164-Cys188 and Cys333-Cys361 in human SHBG), which are highly conserved across species [19]. The four cysteines are also conserved in ursine SHBG and are therefore expected to link Cys167 to Cys192 at the very C terminus of the N-terminal LG domain (Fig. 3) and Cys336 to Cys364 in the C-terminal LG domain.
Human SHBG is a glycoprotein with one O-linked (Thr7) and two N-linked glycosylation sites (Asn351 and Asn367) [27,28]. The O-glycosylation site is fully occupied, while the Asn351 and Asn367 Nglycosylation sites have average molar occupancies of 85% and 95%, respectively [27]. The O-linked glycosylation site is conserved in ursine SHBG (Thr10). The consensus site for N-glycosylation Asn367-Gly-Thr (Asn-X-Ser/Thr, where X is not a Pro) is also conserved in ursine SHBG (Asn370-Gly-Thr), as in most species [29], while Asn351 is substituted to Ser in ursine SHBG (Ser354). However, Pro22 in the human orthologue is substituted with Ser in ursine SHBG (Ser25), giving rise to a new consensus site for ursine N-glycosylation: Asn23-Gly-Ser (Fig. 2). This site is located at the surface of SHBG and is therefore potentially accessible for glycosylation (Fig. 3). The degree of occupancy of these three glycosylation sites in ursine SHBG is not known.
Three bivalent metal ion coordination sites have been identified in the N-terminal LG domain of human SHBG: A Ca 2+ binding site located opposite of the steroid-binding pocket, a Zn 2+ binding site located at the entrance to the steroid-binding pocket, and another Zn 2+ -binding site located close to the dimer interface. The Ca 2+ -binding site and the Zn 2+ -binding site at the dimer interface are conserved in ursine SHBG. For the Zn 2+ -binding site located at the entrance to the steroid-binding site, two of the human Zn 2+ -coordinating residues, His83 and His136, are substituted to Gln in the bear (Gln86 and Gln139, respectively). Therefore, this Zn 2+ -binding site is most likely absent in ursine SHBG. In human SHBG, this Zn 2+ -binding site is located close to a flexible 3 10helical loop (residues 130-135 in human SHBG) that covers the entrance of the steroid-binding pocket when Zn 2+ is absent. Zn 2+ coordination causes a structural change of the loop that decreases the SHBG-binding affinity for estrogens, while this has little or no effect on androgens [30,31]. One of the residues in this flexible lid is substituted in ursine SHBG (Thr132 to Ala135).

Expression of full-length SHBG in ExpiSf cells
To characterize the biochemical properties of SHBG, we established an expression and purification protocol for recombinant ursine and human protein variants. SHBG, fused to a C-terminal FLAG purification tag, was expressed in ExpiSf insect cells from a baculovirus expression vector system. The ExpiSf cell line produced higher SHBG yields compared to test expressions using  Supplementation of 10 µM DHT to the expression medium increased the SHBG yield from ExpiSf cells by 68% for ursine SHBG and 20% for human SHBG (Fig.  S1). Based on preliminary analysis on SHBG expressed in insect cells, we speculate that this might be due to an increased solubility of SHBG upon steroid binding and that the relative effect on increase in solubility is larger for ursine compared to human SHBG (data not shown). However, there might be other explanations, and the effect of adding steroid is not fully understood. The SHBG fusion protein was purified to > 99% by using a combination of anti-FLAG affinity and anion exchange chromatography (AEX), removing all protein contaminants visible by SDS/PAGE with Coomassie R-250 staining (Fig. 4). MALDI-ToF mass spectrometry (MS) confirmed the identity of SHBG, with 89.3% sequence coverage for ursine SHBG and 96.3% for human SHBG (Fig. S2). The final yield (mean AE SD of three replicates) of purified SHBG per liter ExpiSf culture was 65.9 AE 7.5 µg (1.6 AE 0.2 nmol) for ursine SHBG and about 10-fold higher for human SHBG (653 AE 133 µg, 15.8 AE 3.2 nmol).
Ursine and human SHBG contains two signal peptide cleavage sites Cleavage of the signal peptide was evident from western blotting analysis where the apparent molecular weight of SHBG in the conditioned medium was slightly lower than that of SHBG from the cell pellet (Fig. 5). We therefore concluded that SHBG present in the medium had passed through the secretory pathway. Because SHBG is a plasma protein and thus natively secreted from hepatocytes, it is important that the recombinant protein enters the secretory pathway in the host cell to be properly processed, including attachment of carbohydrates and disulfide bond formation [32].
MALDI-ToF MS revealed no peaks corresponding to tryptic or chymotryptic peptides from the signal peptides (human: MESRGPLATSRLLLLLLLLLLRH TRQGWA(L), ursine: MECRGPLATSGWLSLLLLL LPPPGHQ(G)), indicating that the secretion signal peptides had been cleaved off SHBG. A peak corresponding to the most N-terminal semichymotryptic  peptide from mature human SHBG (LRPVLPTQSAH DPPAVHLSNGPGQEPIAVMTF) confirmed cleavage of the signal peptide at the same site as reported previously [33]. We also observed a peak corresponding to an N-terminal semichymotryptic peptide with an alternative cleavage site (RPVLPTQSAHDPPAVHL) for human SHBG. This alternative N-terminal has previously been reported with an occurrence of~25% [33]. For ursine SHBG, two peaks with masses corresponding to N-terminal semichymotryptic peptides with two different signal peptide cleavage sites were apparent, namely between residues 26 and 27 (GRAPRHGL) and between residues 27 and 28 (RAPRHGL) (Fig.  S3). Both cleavage sites are predicted as the most likely by the SignalP algorithm, although this depends on the software version used [34].

Substitution of the secretion signal peptide does not increase SHBG yield
A large fraction of both human and ursine SHBG was retained within the ExpiSf cells compared with the amount secreted to the expression medium (Fig. 5). Furthermore, the amount of ursine SHBG secreted to the expression medium was lower than that of human SHBG, despite similar total expression levels (intraand extracellular) and similar cell densities. This indicates that ursine SHBG is less efficiently secreted from the ExpiSf cells (Fig. 5). In an attempt to optimize secretion efficiency of both ursine and human SHBG from the ExpiSf cells, the native signal peptides were substituted with the signal peptide of honeybee melittin, which has been reported to increase the yield of other proteins from insect cells [35]. Additionally, the signal peptide of ursine SHBG was substituted with the human SHBG signal peptide. Theoretically, the ursine SHBG signal peptide (MECRGPLATSGWL SLLLLLLPPPGHQ(G)) is less efficient than the human (MESRGPLATSRLLLLLLLLLLRHTRQG WA), as it contains a shorter hydrophobic core, fewer polar amino acids in the C-terminal, and several proline residues close to the cleavage site [36,37]. A longer leucine stretch, such as the one found in the human SHBG signal peptide, has been reported to increase the yield of hyenid SHBG in CHO cells [38]. However, substitution of the ursine signal peptide to that of human SHBG did not significantly increase secretion of ursine SHBG from ExpiSf insect cells. Neither did substitution of either the human or ursine signal peptide to that of honeybee melittin (Fig. 6). Thus, an inefficient signal peptide cannot explain the observed low yield of secreted ursine SHBG from the ExpiSf cells.

Recombinant ursine and human SHBG expressed in ExpiSf cells is dimeric
SDS/PAGE analysis of reduced and nonreduced purified recombinant SHBG showed a small change in migration, supporting formation of disulfide bonds in both the ursine and human protein variants (Fig. 7). The presence of the Cys164-Cys188 disulfide bond in human SHBG was verified directly by MALDI-ToF MS of tryptically digested SHBG under non-reducing conditions (Fig. S4). The tryptic dipeptide containing the Cys333-Cys361 disulfide bond in human SHBG, and the dipeptide containing the expected Cys336-Cys364 disulfide bond in ursine SHBG were not observed. These dipeptides both include N-linked glycosylation sites, which may explain their lack of detection. Also, the tryptic dipeptide harboring the predicted ursine SHBG Cys167-Cys192 disulfide was not detected. This is likely due to the relatively large mass (8690.40 Da), which is known to reduce the efficiency of ionization and detection in MALDI-ToF.
In native PAGE, recombinant ursine and human SHBG migrated a similar distance as SHBG present in human serum, suggesting identical multimeric composition (Fig. S5). Indeed, size exclusion chromatographymultiangle light scattering (SEC-MALS) analysis verified the mass of both recombinant and native serum SHBG to be~90 kDa, demonstrating that these proteins form dimers in solution ( Fig. 8; Fig. S6).

Ursine SHBG has three potential glycosylation sites
Ursine SHBG contains the same number of glycosylation consensus sites as the human orthologue, although positioned differently. The masses of SHBG, determined from SEC-MALS analysis (93.1 AE 4.1 for ursine SHBG and 93.7 AE 5.8 kDa for human) demonstrate that the recombinant SHBG is glycosylated, as these masses are larger than the theoretical nonglycosylated dimeric masses (83.1 kDa for ursine SHBG and 82.9 kDa for human). The mass differences correspond to~29 and~32 sugar units per SHBGmonomer for ursine and human SHBG, respectively. Three bands from ursine SHBG and two bands from human SHBG were observed in SDS/PAGE analysis (Fig. 7) and were expected to derive from differential glycan occupancy, as previously described for human plasma SHBG [27]. Moreover, both ursine and human SHBG were observed to elute in more than one peak in AEX chromatography suggesting that various glycoforms occur (data not shown). The masses of recombinant intact monomeric SHBG were determined by MALDI-ToF MS to 43.1 kDa for ursine SHBG and 43.5 kDa for human SHBG (Fig. S7). This corresponds to a mass shift of 1.55 kDa for ursine SHBG and 2.05 kDa for human SHBG, or approximately 9 and 12 sugar units, respectively, relative to the theoretical molecular weight determined from the amino acid sequence (Table 2). Additionally, the peak from human SHBG contained a shoulder, corresponding to a mass of 42.8 kDa and a mass shift of 1.34 kDa or 8 sugar units. Thus, the glycosylations of recombinant human SHBG expressed in insect cells are either fewer or smaller than those of native human SHBG, for which the O-glycosylation typically is 3 sugar units and the two N-glycosylations typically are 11 units each [27]. Since the native glycosylations of SHBG are not associated with steroid-binding, protein folding, or   dimerization, we do not believe that the insect-derived glycosylations affect any of the experimental outcomes in this study [12].
The N-glycosylation consensus site positioned closest to the C terminus of SHBG (Asn367-Gly-Thr in human SHBG) is highly conserved in all known mammalian sequences, including the bear, which suggests that glycosylation of this site in particular is functionally important [29]. We were unable to identify a peak corresponding to the nonglycosylated peptide spanning this N-glycosylation site in ursine SHBG (Asn370), neither in MALDI-ToF analysis of tryptic peptides of our recombinantly expressed SHBG, nor in the data from the bear plasma proteomics study by Welinder et al. [5]. This suggests high glycan occupancy of this site in both native and recombinant ursine SHBG, consistent with the observations for native human SHBG, which has a reported glycan occupancy of 95% [27]. Glycan occupancy results in a mass shift of the peptide, and the resulting modified peptide can only be identified by MALDI-ToF MS if the glycan mass is known.
A peak corresponding to the nonglycosylated peptide spanning the O-glycosylation site in ursine SHBG (Thr10) was observed by MALDI-ToF analysis of recombinant SHBG but not in the proteomics data by Welinder et al. Thus, for recombinant ursine SHBG, the O-glycosylation site is not, or only partially, glycosylated, while it seems to be fully occupied in native ursine SHBG. The O-glycosylation site has been reported to be fully occupied for native human SHBG [27].
A peak corresponding to the nonglycosylated peptide covering the N-glycosylation site Asn23 found in ursine SHBG, but not in human SHBG, was observed both for the recombinant SHBG and in the proteomics data by Welinder et al. [5]. We therefore suggest that this site is not, or only partially, glycosylated. Based on these data, we believe that the size heterogeneity of ursine plasma SHBG observed by Welinder et al. [5] derives primarily from partial occupation of the Asn23 glycosylation site.

Ursine and human recombinant SHBG is active and stabilized by steroid-binding
The ability of the recombinant SHBG to bind steroid was demonstrated by thermal unfolding analysis in the presence and absence of DHT. Supplementation of 50 µM DHT increased the unfolding temperature from 58.7 AE 0.1°C to 64.1 AE 0.3°C for ursine SHBG and from 58.4 AE 0.5°C to 68.0 AE 0.2°C for human SHBG (Fig. 9). Thus, DHT-binding stabilized both ursine and human SHBG, as has previously been reported for human SHBG [31]. Interestingly, human SHBG is stabilized to a larger extend by DHT than ursine SHBG (a 9.6°C versus a 5.4°C increase in unfolding temperature, respectively), suggesting a higher DHT-binding affinity of the human variant.  (Fig. S8). The dissociation constant (K D ) for DHT (mean AE SD) was determined to be 1.32 AE 0.10 nM for ursine SHBG and 0.21 AE 0.04 nM for human SHBG, from triplicate Scatchard plot analysis, as the negative reciprocal of the slope (Fig. 10A). Thus, recombinant ursine SHBG bound DHT with a 6.4-fold lower affinity than human SHBG. A single analysis on a second SHBG batch demonstrated very similar affinities (Fig. 10B). The determined K D value for recombinant human SHBG (0.21 AE 0.04 nM) is in fair agreement to that previously reported for human serum SHBG (0.4 nM) [40].
Next, the SHBG binding affinities for 3βandrostanediol, testosterone, androstenediol, and estradiol were assessed using [ 3 H]DHT as the labeled ligand in a conventional competitive steroid-binding assay. Half maximal inhibitory concentrations (IC 50 ), that is, the steroid concentration able to outcompete 50% of the bound [ 3 H]DHT, were used to determine the relative binding affinities (RBA) to DHT ( Table 3). The RBA values were similar between human and ursine SHBG for these steroids, meaning that their binding affinities to ursine SHBG are also 6-fold lower compared to human SHBG, except for 3β-androstanediol where ursine SHBG demonstrated a slightly but significantly higher relative affinity. In addition, the RBA of both human and ursine SHBG for dehydroepiandrosterone, androstenedione, progesterone, deoxycorticosterone, 11deoxycortisol, and corticosterone were determined to be < 1% of the DHT affinity (data not shown). The IC 50 values for SHBG ligands that effectively displaced DHT from the binding site (i.e., RBA > 1%) were then used to deduce their affinities (K D values) for SHBG by using the Cheng-Prusoff equation (Table 3) [41]. Our in silico analyses of ursine SHBG showed that some of the residues involved in steroid binding differ from those of human SHBG and that substitutions occur within the steroid-binding pocket (Fig. 1). This could explain the lower steroid-binding affinity of ursine when compared to human SHBG. Specific amino acid substitutions, identical to those found in the ursine SHBG steroid-binding pocket, have previously been introduced into human SHBG to determine how particular residues contribute to steroid binding. Introduction of a M107I substitution in human SHBG lowered the DHT-binding affinity about 4-fold with the estradiol-binding affinity being relatively unchanged [42]. Human SHBG with a M139V substitution displayed a 2-to 3-fold reduced affinity for DHT [38]. Since the affinity of ursine SHBG is~6fold lower than human SHBG, we believe that both the M107I and the M139V substitutions contribute to this lowered affinity.
Apart from Old World primates, lower steroid affinity, relative to human SHBG, is a common feature of SHBG in other mammals, including dogs (K D = 7.1 nM [40]) and hyenas (K D = 1.40-1.95 nM [38]), which are both phylogenetically closer to bears than humans. However, the SHBG affinity for the different sex steroids relative to each other, namely DHT > testosterone > estradiol, is conserved between ursine and human SHBG, as in other species [31,43], which suggests that SHBG functions similarly across mammalian species.

Zinc binding does not affect estradiol affinity of ursine SHBG
The relative binding affinity of human and ursine SHBG for estradiol was assessed in the presence and absence of Zn 2+ . It has previously been shown that Zn 2+ coordination has little or no effect on the affinity of human SHBG for androgens [30], and we do therefore not expect the addition of Zn 2+ to affect the binding of [ 3 H]DHT notably. The presence of 1 mM Zn 2+ inhibited the ability of estradiol to outcompete [ 3 H]DHT binding to human SHBG, while Zn 2+ had no effect on the affinity of estradiol toward ursine SHBG (Table 4). Thus, the two ursine His to Gln substitutions in the Zn 2+ coordination site located at the entrance to the steroid-binding pocket eliminates the Zn 2+ -induced reduction in affinity for estradiol observed in human SHBG [26]. The same substitution of the His136 residue has also previously been reported to disrupt the Zn 2+ -binding site in human SHBG [30,44]. The physiological function of the Zn 2+induced reduction in estradiol affinity of human SHBG is not clear, but has been suggested to regulate the availability of sex steroids in Zn 2+ -rich tissues, such as the prostate and other male reproductive organs [44]. Based on our results, we conclude that Zn 2+ does not influence steroid binding of SHBG in bears, as seems to be the case in other mammals apart from humans, since residues within the loop segment covering the steroid-binding pocket are poorly conserved between species, including those critically involved in the binding of Zn 2+ in this location [30].
Our characterization of ursine SHBG will contribute to elucidate the role of SHBG during hibernation in bears In conclusion, we established an insect cell-based expression system for recombinant full-length ursine and human SHBG. This cell system was able to produce dimeric and active full-length ursine and human SHBG. Apart from a 6-fold lower steroid affinity, ursine SHBG had similar characteristics to human SHBG, namely being a homodimeric glycoprotein capable of binding steroids with high affinity. Therefore, ursine SHBG likely has similar biological functions as known for human SHBG. The previous observation of increased SHBG levels in brown bears during hibernation suggests that tight regulation of sex steroid activity is physiologically important [5]. To further delineate the importance of SHBG in bear hibernation, the plasma concentration of SHBG during hibernation and active state should be determined together with the sex steroid concentrations. These data, together with our determined K D of SHBG for various sex steroids, would allow calculation of the free steroid levels. Such information could help determine if sex steroid activity is an important regulatory component of hibernation physiology or if SHBG has other important steroid-free functions.
Further elucidation of the protective mechanisms that underlie the bears' outstanding ability to avoid complications related to immobility and obesity, including the potential involvement of SHBG, may be a stepping-stone to facilitate the development of new human clinical therapies.

Homology model of ursine SHBG
A homology model of the ursine SHBG N-terminal LG domain (PMDB: PM0084113) was generated in the SWISS-MODEL workspace [45] and visualized in PYMOL [46]. The model was based on the crystal structure of the N-terminal LG domain of human SHBG in complex with DHT (PDB: 1D2S) [22]. It should be emphasized that the model of the ursine SHBG Nterminal LG domain does not take energy optimization into account and thus only illustrates the overall structure.

Expression of recombinant SHBG
Liver tissue from a euthanized brown bear was kindly provided by Orsa Rovdjurspark, Sweden. Ethics approval was not required for this study, since no live animals were used, and since euthanasia of the surplus animal was independently decided by Orsa Rovdjurspark. The liver tissue was homogenized by freeze drying followed by bead-beating and the RNA extracted by using the NucleoSpin ® RNA kit (MACHEREY-NAGEL GmbH & Co. KG, Düren, Germany). Ursine and human native SHBG cDNA was generated and amplified by reverse transcription followed by PCR using the ursine RNA and MVP human liver total RNA (Agilent Technologies, Inc., Santa Clara, CA, USA) as template and the primers listed in Table S1. The cDNA was cloned into a pFastBac TM Dual (pFBD) plasmid (Gibco TM , Invitrogen Life Technologies, Carlsbad, CA, USA) downstream to the polyhedrin promoter with a C-terminal FLAG-tag incorporated. The native signal peptide of human and ursine SHBG was substituted with that of honeybee melittin (MKFLVNVALVFMVVYISYIYAAPEP) and for ursine SHBG also with that of human SHBG (MESRGPLATSRLLLLLLLLLLRHTRQGWA) by PCR using the pFBD plasmids as templates, whereafter the SHBG constructs were cloned into pFBD as above. The pFBD plasmids were transformed into MAX Efficiency ® DH10Bac TM Competent E. coli cells (Gibco TM , Invitrogen Life Technologies) for sitespecific transposition of SHBG into the bacmid vector according to the method described in the user manual 'Bac-to-Bac ® Baculovirus Expression System' (version D, 6 April 2004, 10359) from Invitrogen TM (Life Technologies, Carlsbad, CA, USA). Correct transposition of the SHBG constructs into the bacmids was verified by sequencing. The insect cell line ExpiSf9 TM (Gibco TM , Invitrogen Life Technologies) was transfected and cultivated according to the manufacturer's instructions ('ExpiSf TM Expression System', Publication Number MAN0017532, Revision A.0). Briefly, ExpiSf9 TM cells were cultured in ExpiSf CD medium (Gibco TM , Invitrogen Life Technologies) supplemented with 1% PenStrep (100 unitsÁmL −1 penicillin, 100 gÁmL −1 streptomycin) (Gibco TM , Invitrogen Life Technologies) at 27°C with normal atmosphere and 95 rpm orbital shaking (50-mm shaking diameter). The ExpiSf9 TM cells were transfected with the bacmid DNA with use of ExpiFectamine TM Sf Transfection Reagent (Gibco TM , Invitrogen Life Technologies) to generate recombinant baculovirus which was further amplified and harvested when the viability dropped to 60-80%. The baculoviral titer was determined by a flow cytometric tittering assay using the Baculovirus Envelope gp64 Monoclonal Antibody (AcV1), PE, eBioscience TM (Invitrogen TM , Life Technologies). Only viral stocks with titers ≥ 1 × 10 8 ivpÁmL −1 were applied for protein expression. For protein expression, ExpiSf cells were seeded on the day before infection to a final density of 5 × 10 6 viable cellsÁmL −1 in fresh ExpiSf TM CD Medium and added ExpiSf TM Enhancer (Gibco TM , Invitrogen Life Technologies). After 18-24 h, the cells were infected with a multiplicity of infection ≥ 5. The expression medium was harvested 72 h postinfection where cells were removed by centrifugation at 500 rcf for 15 min (RT) and remaining cell debris by centrifuging at 18 600 g for 30 min (4°C). 1 mM CaCl 2 and 0.02% sodium azide were added to stabilize the protein and prevent microbial growth.

Protein purification
Ursine and human SHBG with native signal peptide was purified from 1 L cultures by anti-FLAG affinity chromatography and anion exchange chromatography.
1 mL (settled) equilibrated Pierce TM Anti-DYKDDDDK Affinity Resin (Pierce Biotechnology, Inc., Thermo Fisher Scientific, Rockford, IL, USA) was added per liter expression medium and incubated overnight on a roller mixer at 4°C. The resin was collected from the expression medium by centrifugation at 1000 g for 5 min and thereafter transferred to an empty gravity flow column. The resin was washed with 100 mL binding buffer (50 mM Tris-HCl, 150 mM NaCl, 2.5 mM CaCl 2 , pH 7.5). Twice during the wash step, the column was plugged and capped, and the column mixed over end with buffer, before eluting.
FLAG-tagged SHBG was eluted from the resin by 30 min incubation with 2 mL FLAG elution buffer [binding buffer supplemented with 100 μgÁmL −1 FLAG ® Peptide (Sigma-Aldrich, Saint Louis, MO, USA)] in the capped column at 4°C with over end mixing. The elution step was repeated with 15 min incubation. Remaining unbound SHBG was eluted with 2 × 2 mL binding buffer.
The anti-FLAG purified SHBG was additionally purified by anion exchange. The pooled sample was mixed with two volumes of dilution buffer (50 mM Tris-HCl, 2.5 mM CaCl 2 , pH 7.5), to lower the salt concentration, and loaded on a Mono Q HR 5/5 column (Pharmacia Biotech, Uppsala, Sweden) preequilibrated in buffer A (50 mM Tris-HCl, 2.5 mM CaCl 2 , 50 mM NaCl, pH 7.5). The column was washed with 10 column volumes of buffer A, whereafter proteins were eluted by applying a linear salt gradient from 0% to 100% buffer B (50 mM Tris-HCl, pH 7.5, 500 mM NaCl, 2.5 mM CaCl 2 ) over 50 column volumes at a flow rate of 0.5 mLÁmin −1 . 1 mL fractions were collected and those containing SHBG were combined.
The identity, purity, and integrity of the SHBG proteins were confirmed by SDS/PAGE, western blotting, and MALDI-ToF MS. The purified protein yield from three independent culture batches was determined using absorption at 280 nm (extinction coefficients for the recombinant ursine and human SHBG are 47230 M −1 Ácm −1 and 58230 M −1 Ácm −1 , respectively).

Test expression of recombinant SHBG in ExpiSf cells supplemented with DHT
Test expressions were performed in 4 mL suspension cultures in 6-well plates (CELLSTAR ® , Greiner Bio-One GmbH, Kremsmünster, Austria). ExpiSf cells were seeded, infected, and harvested as previously described, except that DHT dissolved in absolute ethanol was supplemented to the cultures at the time of infection, resulting in final DHT concentrations of 0 and 10 µM, and an ethanol concentration of 1% in the cultures (ethanol in this concentration did not affect the morphology of the cells).

Western blot analysis
The proteins were blotted from an SDS/PAGE (4-20%, except for native PAGE where a 10% gel was applied) onto a nitrocellulose membrane using a Trans-Blot ® Turbo TM Transfer System (Bio-Rad Laboratories, Singapore).
For comparison of intracellular and extracellular SHBG, cells and expression medium were separated by centrifugation at 300 g for 10 min, and the cells were resuspended in PBS corresponding to the original volume. Equal volumes of intra-and extracellular samples were loaded on the gel.
For immunostaining of recombinant SHBG in a native western blot, affinity purified rabbit anti-SHBG LG4 primary antibody (provided by Geoffrey Hammond, The University of British Columbia) and Goat anti-rabbit IgG (H + L), HRP secondary antibody (Invitrogen TM , Life Technologies) was applied and detected by Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare, Buckinghamshire, UK).

Sample preparation for digested SHBG
Tryptically digested SHBG was analyzed at both reducing and nonreducing conditions and chymotryptically digested SHBG at reducing conditions. Purified SHBG was mixed 1 : 1 (v/v) with digestion buffer [2% (w/v) sodium deoxycholate in 100 mM triethylammonium bicarbonate] and incubated 10 min at 99°C. For the samples to be reduced, tris(2carboxyethyl)phosphine was added in a 1 : 25 (w/w) reagent-to-protein ratio after cooling below 37°C and incubated 30 min at 37°C, whereafter iodoacetamide was added in a 1 : 10 (w/w) ratio and the sample incubated for 20 min at 37°C in the dark. Trypsin or chymotrypsin was added in a 1 : 50 (w/w) ratio and the samples incubated overnight at 37°C.
To precipitate sodium deoxycholate, formic acid was added to a concentration of 2.0% and the sample incubated at room temperature for 5 min. The samples were then centrifuged at 13 000 g for 20 min at 4°C. The supernatant was desalted in a column with a C18 disc (Empore TM , Oxford, PA, USA) and R2 and R3 Poros beads (Applied Biosystems, Foster City, CA, USA) and eluted with 10 µL 80% acetonitrile and 0.1% formic acid.

Sample preparation for intact SHBG
The method applied is described by Signor and Erba [48].
Purified SHBG was desalted in a column with a C8 disc (Empore TM ) and R2 and R3 reversed phase resin and eluted with 10 µL 80% acetonitrile and 0.1% formic acid. A small amount of saturated α-CHCA acetone solution (Thermo Fisher) was deposited on an AnchorChip TM MALDI target plate to form a matrix thin layer. 0.5 μL desalted protein [or protein standard II (Bruker Daltonik)] was deposited on the α-CHCA thin layer followed by 0. MALDI-ToF MS was performed using a Bruker Daltonics autoFlex TOF/TOF MALDI-ToF instrument. Spectra were obtained in positive reflector mode for peptides and in positive linear mode for intact protein.
Size exclusion chromatography with multiangle light scattering analysis SEC-MALS was used to evaluate the molecular masses of SHBG. SEC measurements were performed on an Agilent 1260 Infinity HPLC system in-line with a WYATT 30S5 guard column and a WYATT 30S5 SEC column. A MALS detector DAWN HELEOS 8 (Wyatt Technology, Santa Barbara, CA, USA) was used to obtain protein molecular mass and the refractometer Optilab T-rEX (Wyatt Technology) to measure the refractive index (dRI). The system was equilibrated overnight in 20 mM Tris-HCl, 150 mM NaCl, 1 mM CaCl 2 , 1 μM DHT, pH 8.0 (0.22 μm filtered and sonicated). For the ursine measurements, 250 mM NaCl was applied to prevent the protein from sticking to each other. 100 μL sample was injected (~10 μg SHBG) and BSA (2 mgÁmL −1 ) was used for calibration. The flow rate was 0.5 mLÁmin −1 . Data analysis was performed using the ASTRA ® software (Wyatt Technology).

Thermal unfolding profiles
Five hundred nanomolar purified ursine and human SHBG was equilibrated with 0 or 50 μM DHT in triplicates. Thermal unfolding profiles from 20.0 to 95.0°C and a temperature increase of 1.0°CÁmin −1 were measured on a Prometheus NT.48 system (NanoTemper Technologies, München, Germany) using highsensitivity capillaries. The unfolding temperatures were determined as the peak maximum of the first derivative of the fluorescent ratio 350 nm/330 nm as a function of the temperature.

Steroid binding assays
The binding affinities and relative binding affinities (RBA) of ursine and human SHBG were measured by steroid ligand saturation analysis according to the method by Hammond and Lähteenmäki [39].
The optimal dextran-coated charcoal (DCC) exposure time was determined to 10 min by time-course experiments, which ensured efficient adsorption of the majority of the nonbound [1,2-3 H]5α-dihydrotestosterone ([ 3 H]DHT) (specific activity 60 CiÁmmol −1 , American Radiolabelled Chemicals) with minimal dissociation of bound complex (Fig. S9A). The percentage dissociation of bound complex during the 10 min DCC exposure was determined by extrapolation of the counts from specifically bound [ 3 H]DHT to SHBG to zero time and used to calculate the dissociation rate of the complex (Fig. S9B,C). After 10-min DCC exposure, approximately 23% of the ursine SHBG-DHT complex and 6% of the human SHBG-DHT complex had irreversibly dissociated. Consequently, a correction factor of 1.23 for ursine SHBG and 1.06 for human SHBG was applied in the calculations of SHBG binding capacity to account for dissociation during the DCC treatment.

Determination of SHBG-binding affinities to DHT
The steroid-binding affinities of purified recombinant ursine and human SHBG to [ 3 H]DHT were determined by equilibration with [ 3 H]DHT in 7-step twofold dilutions in the presence or absence of 100-fold molar excess of unlabeled DHT (Steraloids, Inc., Newport, RI, USA) and removal of unbound steroid by 10-min DCC exposure. For Scatchard analysis, bound steroid over free steroid was plotted as a function of bound steroid. The free steroid count is calculated by subtracting bound steroid count from the total [ 3 H] DHT count added. Linear regression analysis was performed to obtain the dissociation constant (K D ), corresponding to the negative reciprocal of the slope.

Determination of relative binding affinities
The RBAs were determined in a competitive steroidbinding assay, by saturating SHBG with [ 3 H]DHT (approximately 6 nM) and adding increasing amounts of the following unlabeled steroid competitors: estradiol, testosterone, androstenediol, 3β-androstanediol, dehydroepiandrosterone (Aldrich Chemical Company, Inc., Milwaukee, WI, USA), androstenedione (Sigma Chemical Co., St. Louis, MO, USA), progesterone, 11deoxycorticosterone, 11-deoxycortisol, and corticosterone. Additionally, the RBA for estradiol was assessed in the presence and absence of 1 mM Zn 2+ . Steroids were purchased from Steraloids, Inc., unless stated otherwise. RBA were calculated in percentage from the IC 50 values, that is, the concentrations of steroid that result in 50% displacement of the bound [ 3 H]DHT relative to the concentration required of unlabeled DHT.
K D values of the steroids were calculated from the IC 50 values, according to the Cheng-Prusoff equation [41]:

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Effect of DHT supplementation on the SHBG yield. Fig. S2. Sequence coverage of SHBG in MALDI-ToF MS. Fig. S3. Identification of the signal peptide cleavage site of ursine SHBG. Fig. S4. Identification of disulfide bond in human SHBG. Fig. S5. Native western blot of SHBG. Fig. S6. Chromatograms from SEC-MALS. Fig. S7. Determination of SHBG molecular mass by MALDI-ToF MS. Fig. S8. Saturation curve of DHT binding to SHBG. Fig. S9. Time-course experiments during DCC exposure. Table S1. Primer sequences used for cloning.