Identification of Lysine 134 in the Steroid-binding Site of the Sex Steroid-binding Protein of Human Plasma *

The sex steroid-binding protein of human plasma SBP (or sex hormone-binding globulin, SHBG) was specifically inhibited with the alkylating affinity label, 178-[([2-14C]bromoacetyl)oxy]-5a-androstan-3-one. The natural ligand, 5a-dihydrotestosterone, was shown to protect against inactivation and labeling. The steroid-binding activity of the protein was abolished when approximately 1 mol of label was incorporated into 1 mol of dimeric SBP. In order to identify and locate the labeled amino acid in the steroid-binding site, the steroidal portion of the bound label was first removed and the protein was digested with Achromobatter protease and subdigested with trypsin. Seven radioactive peptides were isolated, sequenced, and found to contain the common sequence QVSGPLTSXR. Residue X was identified as lysine-134 from the SBP amino acid sequence (Walsh, K. A., Titani, K., Kumar, S., Hayes, R., and Petra, P. H. (1986) Biochemistry 25,7584-7590). The results indicate that only 1 of the 2 lysine-134 residues in the homodimer was labeled. This suggests that the steroid-binding site is constructed from an association of the two subunits in an AB to BA “sandwich” configuration with lysine-134 residue of one subunit on one surface near the D-ring and the lysine-134 of the other subunit at the opposite end of the steroid, or well away from the steroidbinding site. Although the nature of the data does not allow description of a specific role for lysine-134, its proximity to the 17&OH of the steroid nucleus suggests participation in the binding process through direct or indirect hydrogen bonding.

The sex steroid-binding protein of human plasma SBP (or sex hormone-binding globulin, SHBG) was specifically inhibited with the alkylating affinity label, 178-[([2-14C]bromoacetyl)oxy]-5a-androstan-3-one. The natural ligand, 5a-dihydrotestosterone, was shown to protect against inactivation and labeling. The steroid-binding activity of the protein was abolished when approximately 1 mol of label was incorporated into 1 mol of dimeric SBP. In order to identify and locate the labeled amino acid in the steroid-binding site, the steroidal portion of the bound label was first removed and the protein was digested with Achromobatter protease and subdigested with trypsin. Seven radioactive peptides were isolated, sequenced, and found to contain the common sequence QVSGPLTSXR. Residue X was identified as lysine-134 from the SBP amino acid sequence (Walsh, K. A., Titani, K., Kumar, S., Hayes, R., and  Biochemistry 25,7584-7590).
The results indicate that only 1 of the 2 lysine-134 residues in the homodimer was labeled. This suggests that the steroid-binding site is constructed from an association of the two subunits in an AB to BA "sandwich" configuration with lysine-134 residue of one subunit on one surface near the D-ring and the lysine-134 of the other subunit at the opposite end of the steroid, or well away from the steroidbinding site. Although the nature of the data does not allow description of a specific role for lysine-134, its proximity to the 17&OH of the steroid nucleus suggests participation in the binding process through direct or indirect hydrogen bonding.
The pure protein was concentrated to l-2 mg/ml, dialyzed against 10 mM Tris-Cl, 10 mM NaCl, 5 mM Caf&, and stored at -20 "C. The levels of DHT remaining were found to be very low as estimated by radioimmunoassay. Pure human SBP is stable in the absence of DHT and glycerol and can be stored for months at -20 "C without loss of activity. The concentration of SBP was determined spectrophotometrically using eZRO = 1.14 X lo5 cm-' M-' (Petra et al.,i986a)and M, = 93,400 (Petra et al., 198613). Assay of SBP Actiuity-The DEAE-cellulose filter assay previously described for measuring steroid-binding proteins in plasma (Mickelson and Petra, 1974;Schiller and Petra, 1975) was used with some modification.
Aliauots from the DHTBr reaction vessel (20 ~11 containing about 2 bg of SBP) were added to 200 ~1 of 4% bovine serum albumin solution followed by a 30-min incubation at 25 "C. These were then added to 1.5.mg pellets obtained from 0.5% charcoal, 0.05% dextran (w/w), 0.1% gelatin, 10 mM Tris-Cl, pH 7.4. The suspensions were shaken gently for 15 min at room temperature and centrifuged at 4 "C for 5 min at 18,000 X g to remove charcoal-containing steroid. Aliquots were further diluted lo-fold with 10 mM Tris-Cl, pH 7.4, and incubated for 20 min at room temperature with a S-fold molar excess of ["HIDHT over SBP, in the presence or absence of 100 times molar excess of radioinert DHT over ["HIDHT. The solutions were further incubated for 20 min at 0 "C, and 50.~1 aliquots were applied to a stack of two DEAE-filter paper discs for assay. Synthesis of 17b-f(Bromoacetyl)oxy]-5n-androstan-3-one-Synthesis was carried out according to published procedures (Sweet and Samant, 1980) starting with 0.1 mmol of DHT, bromoacetic acid, and 1,3-dicvclohexylcarbodiimide.
After mixing the reactants in 4 ml of anhydrous CH&l,, the reaction was proceeded under dry N2 for 4.5 h at 0 "C. The product was purified on Silica Gel G (6 g, 0.9. X locm column) collecting 0. The solution was adjusted to pH 11 and incubated at 37 "C for 1 h, adjusted to pH 7.4, dialyzed overnight at 4 "C against 10 mM Tris-Cl, pH 7.4, and lyophilized. The protein was then reduced and S-carboxymethylated in 1 ml of 6 M guanidine HCl as described by Takio et al. (1983). Gel Electrophoresis-SDS-gel electrophoresis was carried out according to Petra et al. (1986a). Native gel electrophoresis was carried out in tubes according to Petra et al. (1983). Amino Acid Analysis and Peptide Sequence Determination-Amino acid analyses were carried out as previously published (Bidlingmeyer et al., 1984). Peptides were sequenced with an Applied Biosystems model 470 Sequencer with on-line phenylthiohydantoin analysis using published programs (Hunkapiller et al., 1983).

Mass Spectrometric
Analyses-Time-of-flight measurements were carried out by Dr. Pat Griffin at Genetech (San Fransisco, CA) according to published procedures (Griffin et al., 1989).

Kinetics of DHTBr
Labeling-The kinetics of SBP inactivation with DHTBr and incorporation of the label are shown in Fig. 1, A  normally migrates, and peak II (minor) at about 28,000. The presence of DHT in the mixture lowers the radioactivity content of both peaks. The SDS-PAGE data of Fig. 1B indicate that about 0.8 mol of label is incorporated per mol of dimeric SBP after 4 h; when peak II is added to peak I, the stoichiometry reaches about 1.1 mol/mol. A stained SDS slabgel containing samples from a MO-min reaction and from unreacted SBP is also shown in Fig. 2. As expected, there are two stained bands at about 44,000 and 28,000 corresponding to radioactive peaks I and II (lane 3). Native SBP (lane 2) does not contain the M, = 28,000 band.
Electrophoresis of Labeled SBP in Native Gels-In order to determine whether or not labeling disrupts the dimeric structure of SBP, protein was reacted with DHTBr as described above and electrophoresed in native gels along with native SBP. Fig. 3  same place as native protein indicating that the native dimeric structure is maintained.

Removal of Steroid from Labeled SBP-To gain further
insight into the labeling reaction, SBP was inactivated with ['C]DHTBr as described above, and also with [1,2-"H] DHTBr, which contains the label in the steroid nucleus instead of the 17/3 side chain. As described above, both reactions were exhaustively dialyzed against IAA, and against water to remove IAA. Both reactions were then incubated at pH 11 for 1 h at 37 "C to cleave the covalently bound steroid through hydrolysis of the ester linkage in the acetoxy side chain. The freeze-dried products were electrophoresed in SDS. The gel pattern shown in Fig. 4A demonstrates that the 3Hlabeled steroid was removed since neither peak I nor II was radioactive after base treatment, whereas SBP retained label from [W]DHTBr (Fig. 4B). The data further suggest that the label is not esterified to aspartic or glutamic acid residues because the '"C label remains with the protein at high pH. These experiments show that cleavage of the acetoxy ester bond by base removes the steroidal moiety of the label, thereby reducing the hydrophobicity of labeled peptides and facilitating their isolation.
Identification of the Site of Affinity Labeling-Fifteen nmol of the affinity-labeled SBP polypeptide (calculated as the monomer, M, = 46,700), was dialyzed as described above, incubated at pH 11 to remove the steroidal moiety, and reduced and S-carboxymethylated.
After dialysis against 0.1 M NH,HCO:, the sample contained 50,000 cpm. Since 15 nmol (monomeric) of labeled SBP should contain about 18,000 cpm based on label incorporation of 1.1 mol/mol of dimeric SBP (combined peaks I and II of Fig. 2), the data indicated that about 32,000 cpm of unbound label was still present after dialysis. To reduce protein losses, the labeled protein was not dialyzed further but digested at lysyl residues and fractionated by size (see "Experimental Procedures"). Two peptide-bound radioactive fractions were obtained, and each was fractionated on reverse phase HPLC C3 columns to yield four radioactive fractions containing a total of 5,275 cpm (2.1 nmol of label, 25% yield). Preliminary analyses of small aliquots suggested that these fractions contained overlapping products of incom- subdigested with trypsin, and products were separated on a C8 reverse phase HPLC column. As shown in Fig. 5, five labeled peptide fractions were recovered and 4080% of each was subjected to Edman degradation. The results in Table I reveal that the common sequence QVSGPLTSXR is present in each fraction, although minor contaminants are seen as well as long versions of the same peptide. The residue denoted X in the major peptides of Table   I did not yield an identi~able phen~lthioh~dantoin but, from the amino acid sequence of human SBP (Walsh et al., 1986), that residue corresponds to lysine 134. Since all the other residues in this sequence were easily recognized, it is concluded that Lys-134 is the site of ~-carboxymethylation by the affinity label, and that the phenylthiohydantoin of N'carboxymethyllysine does not elute in a readily recognized location in our chromatographic system.
Assignment of Lys-134 as the site of labeling was confirmed by time-of-flight mass spectrometry on fractions T-17 and T-20 of Table I. The major ion in T-17 had a m/z of 1400.6, corresponding to the calculated mass (1400.5) of N'-carboxy-methyGLRQVSGPLTSKR. Fraction T-26 reveaied two ions, of m/z 1749 and 1766. The difference in mass between these two ions corresponds to the 17 mass units of lost NH3 when glutamine cyclizes to a pyroglutamyl residue. Therefore, the two ions in T-20 correspond in mass to the ~-carboxymethyl forms of QVSGPLTSKRHPI~R and its blocked pyrogluta-my1 derivative. The blocked pyroglutamyl form of that peptide probably accounts for much of the label in T-22, although this was not proven by mass spectrometry.
In summary, all of these results point to Lys-134 as the site of labeling. No other minor peptide was found in more than one labeled fraction. As shown in Table I, the specific activity of the label (2500 cpm/nmol) agrees well with the amount of the major peptide observed in the Sequencer with each of the peptide preparations. The total amount of radioactivity recovered as Lys-134~containing peptides was 15-20%; this is a good recovery considering that the generally accepted peptide recovery yield per HPLC column is 50%.

DISCUSSION
Description of the chemical environment of steroid binding sites is needed for understanding the molecular basis of steroid-binding specificity. In this paper we report the presence of lysine 134 in the steroid-binding site of human SBP by virtue of its specific alkylation with the affinity label, 17/3-[ (bromoacetyl)oxy~-5ff-androstan-3-one.
Specificity of the labeling reaction is shown by protection with the natural ligand, 5ff-dihydrotestosterone. Since inco~oration of approximately 1 mol of label per mol of dimeric SBP abolishes the steroid-binding activity of the protein, ianctivation is a direct consequence of Lys-I34 labeling.
As shown in Fig. 1, inactivation and label inco~oration also occur in the presence of DHT, although at a lower rate. Incomplete protection would occur if DHTBr were to compete efficiently with DHT, thereby displacing it in the steroidbinding site. This would result in a gradual decrease in SBP activity with time as indicated in Fig. 1, Similar observations were reported for DHTBr inactivation of the androgen receptor in prostate cytosols where protection by DHT was also found to be incomplete (Kovacks and Turney, 1988). These authors report that DHTBr binding to the receptor was only 2.3-fold lower than DHT. We have not analyzed peptides obtained from labeled SBP in the presence of DHT, but we strongly suspect that lysine 134 is also alkylated under those conditions. This is supported by the fact that only one site of modification was found when the protein was reacted in the absence of DHT. Although similar data would be expected if some labeling were to occur at nonspecific sites in addition to the steroid-binding site, this seems unlikely because label incorporation and Ioss of activity occur at the same rate in a 1:l stoichiometry in the absence of DHT. Therefore, the likely explanation is that DHTBr competes efficiently with DHT for the binding site. Fig, 1 also indicates that the rate of inhibition and label Although we did not carry out a complete kinetic analysis, this phenomenon is probably caused by a decrease in DHTBr concentration as reaction proceeds. This effect is less pronounced in the absence of DHT because less affinity label is required to maintain a pseudo-first order reaction. In the presence of DHT, however, more reagent is needed to maintain the rate of the reaction which is dependent upon efficient competition with DHT. The data also indicate that labeling of only one of the two identical subunits of homodimeric SBP is sufficient to abolish the activity. The radiolabeled peptides contain only one site of modification, carboxymethyllysine 134. Since there are two Lys-134 residues per dimer, the second one must remain unmodified during the inhibition reaction and therefore may not be located near the D-ring of the steroid. This suggests that the steroid-binding site is constructed through an association of the two subunits in an AB to BA "sandwich" configuration.
This interpretation lends support to the hypothesis placing the steroid at the interface between the subunits (Petra et al., 1983). This molecular arrangement would provide an explanation for the problem of a homodimerit protein molecule recognizing an asymmetric ligand. The AB/BA configuration would present two different surfaces to the two faces of the steroid. The results presented here support the positioning of the Lys-134 of one subunit on one surface near the D-ring, while the Lys-134 of the other subunit could be either at the opposite end of the steroid or well away from the steroid-binding site. Analysis of SDS gels reveals that in addition to the M, = 44,000 labeled SBP band there is a second radioactive band at about 28,000 (Fig. 2). The origin of this band remains unclear. It does not represent a contaminating protein because it is absent from the SDS-PAGE pattern of the unreacted protein (Fig. 2, lane 2). It appears that labeling at Lys-134 may promote a side reaction involving cleavage of the SBP polypeptide chain. In any case, this side reaction does not affect our interpretation of the data since the digested material from which the labeled peptides were isolated contained both M, = 44,000 and 28,000 components and all the radioactive peptides that we isolated contained the sequence encompassing Lys-134.
DHTBr has the characteristics of an ideal affinity label for steroid-binding proteins and points the way to the design of isomeric labels with the reactive group at different positions. Placing the radioactive isotope in an acetoxy side chain is particularly advantageous because it provides an experimental means for removing the steroid while leaving the radioactive tag covalently attached to an amino acid residue. Removal of the hydrophobic steroid facilitates isolation of steroid-binding site peptides in good yield. The results presented here also show that the conditions used for removing the steroidal moiety are mild enough to preserve the primary structure of the protein.
Although the nature of the data does not allow us to describe a specific role for Lys-134, its proximity to the 17&OH of the steroid nucleus nevertheless suggests participation in the binding process through two possible mechanisms. First, the e-amino group of Lys-134 could form a hydrogen bond with the 17/3-OH group providing energy of binding for stabilizing the steroid in the binding site. Presence of the hydrogen on the 17P-OH of dihydrotestosterone or testosterone is required for high affinity; derivatives that contain a ketone at that position or a 170i-OH group do not bind to SBP. This proposed mechanism would require that the steroid-binding site Lys-134 be mostly un-ionized at pH 7 and therefore have an unusually low pK for a primary amino group. The hydrophobic environment of the steroid-binding site could perhaps provide a medium of low dielectric constant necessary to maintain this low pK. The other mechanism would involve Lys-134 in a salt bridge with a nearby carboxyl group which in turn would form a hydrogen bond with the 17/3-OH group. Hydrogen bond stabilization of this kind has been documented in the active site of other proteins and is analogous to the chargerelay system postulated to exist in the active site of trypsin and other "serine" proteases.
In conclusion, using the classical approach of affinity labeling, we have presented evidence that 1 of the 2 Lys-134 residues of the dimer resides in the steroid-binding site. We postulate that this residue may be involved in the steroidbinding process by forming a hydrogen bond with the 17/3-OH group of the steroid either directly or indirectly.
A recent report suggests that Met-139 is modified by the photoaffinity label, A6-testosterone, which contains its reactive group at the 3@-position of the steroid (Grenot et al., 1988). It is therefore conceivable that the portion of the polypeptide chain encompassing these 2 residues (-PLTSKRHPIMRIAL-) may have significant contact with a region of the steroid between the D-ring and the A-ring and may thus play an important role in the construction of the steroid-binding site. We had previously hypothesized that the alternating leucine segment located between residues 247 and 291 might represent part of the steroid binding site due to its hydrophobicity (Petra et al., 1988). Although we do not rule out that possibility, the present results indicate direct involvement of a different part of the protein.  . Since a N'carboxymethyllysine standard was not included in their analyses of alkylated protein hydrolysates, alkylation of a lysine residue cannot be ruled out in that work. Furthermore, Edman degradation of their radioactive peptide preparation yields phenylthiohydantoin-histidine at position 235 in relatively good yield for a charged amino acid indicating that His-235 is not alkylated.