Crystal Structure of Human Cholesterol Sulfotransferase (SULT2B1b) in the Presence of Pregnenolone and 3′-Phosphoadenosine 5′-Phosphate

The gene for human hydroxysteroid sulfotransferase (SULT2B1) encodes two peptides, SULT2B1a and SULT2B1b, that differ only at their amino termini. SULT2B1b has a predilection for cholesterol but is also capable of sulfonating pregnenolone, whereas SULT2B1a preferentially sulfonates pregnenolone and only minimally sulfonates cholesterol. We have determined the crystal structure of SULT2B1a and SULT2B1b bound to the substrate donor product 3′-phosphoadenosine 5′-phosphate at 2.9 and 2.4 Å, respectively, as well as SULT2B1b in the presence of the acceptor substrate pregnenolone at 2.3 Å. These structures reveal a different catalytic binding orientation for the substrate from a previously determined structure of hydroxysteroid sulfotransferase (SULT2A1) binding dehydroepiandrosterone. In addition, the amino-terminal helix comprising residues Asp19 to Lys26, which determines the specificity difference between the SULT2B1 isoforms, becomes ordered upon pregnenolone binding, covering the substrate binding pocket.

Sulfonation of endogenous compounds and xenobiotics is catalyzed by a large group of enzymes called sulfotransferases. These enzymes catalyze the transfer of a SO 3 group from PAPS, 1 the universal sulfonate donor molecule, to an acceptor group of various substrates. Cytosolic sulfotransferases sulfonate small molecules such as hormones, neurotransmitters, bioamines, and therapeutic drugs (1,2). Hydroxysteroid sulfotransferases sulfonate the 3-hydroxy group of steroids and are part of the superfamily of cytosolic sulfotransferases that includes phenol (aryl) and amine sulfotransferases. The prototypical human hydroxysteroid sulfotransferase SULT2A1 readily sulfonates dehydroepiandrosterone (DHEA) but also has a broad substrate predilection (3). Recently, two human hydroxysteroid sulfotransferases, SULT2B1a and SULT2B1b, have been shown to be encoded by the same gene but differ at the amino terminus by 8 and 23 amino acids, respectively, as a result of an alternative exon 1 (4). The SULT2B1a isoform avidly sulfonates pregnenolone, whereas sulfonation of cholesterol is minimal. In contrast, the SULT2B1b isoform preferentially sulfonates cholesterol with greater efficiency and, to a lesser extent, pregnenolone (5). Although both isoforms are capable of sulfonating DHEA, they do so with relatively low efficiencies (3).
Differential expression patterns of SULT2B1a and SULT2B1b in organ systems, particularly the skin and brain, in association with their respective substrate preferences reveal potential physiological implications for the sulfonated product. SULT2B1b, now recognized as a cholesterol sulfotransferase, is quantitatively the predominant hydroxysteroid sulfotransferase expressed in human skin. Cholesterol sulfate has been recognized to be essential in skin development as a regulatory molecule in human keratinocyte differentiation and creation of the barrier (6 -9). The human fetal brain appears to only express the SULT2B1a isoform (10), which is consistent with the evidence that the brain and spinal cord in mouse almost exclusively express SULT2B1a (11). Pregnenolone sulfate, which is most efficiently produced by SULT2B1a, is now acknowledged as an essential neurosteroid that modulates neurotransmitters such as ␥-aminobutyric acid type A, N-methyl-D-aspartate, and Sigma 1 (11)(12)(13)(14)(15)(16).
Although the SULT2A1 and SULT2B1 isozymes are ϳ37% identical in amino acid sequence, the SULT2B1 isoforms have extended amino-and carboxyl-terminal ends that are absent in the SULT2A1 isozyme (5). Previous studies showed that removal of the 52 amino acid carboxyl-terminal end that is common to both SULT2B1 isoforms has no effect on catalytic activity of either isoform (5). Removal of the 8-residue aminoterminal end that is unique to SULT2B1a has no significant effect on catalytic activity; however, removal of the 23-residue amino-terminal end that is unique to SULT2B1b abolishes catalytic activity for cholesterol but not pregnenolone (5). Full wild-type activity of SULT2B1b for cholesterol can be retained if only the first 18 amino acids are deleted (5). This finding suggests that residues 19 DISEI 23 are responsible for the ability of SULT2B1b to sulfonate cholesterol. Thus, the difference in substrate specificity appears to lie at the unique amino terminus of the SULT2B1 isoforms.
To better understand the nature of the substrate specificity differences among SULT2A1, SULT2B1a, and SULT2B1b, we have solved the x-ray crystal structures of SULT2B1a and an active truncated version of SULT2B1b (Asp 19 -Asp 312 ) in the presence of PAP, the substrate donor product, as well as the structure of SULT2B1b in the presence of PAP and the acceptor substrate pregnenolone. These structures reveal a different orientation of acceptor substrate binding for SULT2A1 and the SULT2B1 isoforms and suggest a possible role for the unique amino terminus of SULT2B1b in determining substrate specificity for cholesterol.
SULT2B1a fusion protein was expressed overnight in 2ϫ YT (16 g of bacto-tryptone, 10 g of bacto-yeast extract, 5 g of NaCl, and 5 ml of 1 M NaOH in 1 liter) media at 22°C on a shaker after induction with isopropyl-1-thio-␤-D-galactopyranoside at a final concentration of 0.2 mM. Cells were pelleted, resuspended in the lysis buffer (50 mM Hepes, pH 7.5, and 350 mM NaCl), sonicated on ice, and spun down. Soluble protein was loaded onto glutathione-Sepharose 4B resin. The resin was washed extensively in batch with lysis buffer plus 1 mM EDTA and 1 mM dithiothreitol. The full-length SULT2B1a enzyme was eluted from the resin by PreScission protease (Amersham Biosciences) digestion, and the protein was concentrated to 15 mg/ml and then dialyzed overnight against 20 mM Hepes, pH 7.5, and 100 mM NaCl. Protein was loaded onto a PAP-agarose column but did not bind. The flow-through was then loaded onto a Q-Sepharose column and eluted with a salt gradient from 100 mM to 1.0 M NaCl. The eluted protein was then concentrated to 13 mg/ml and dialyzed against 20 mM Hepes, pH 7.5, and 100 mM NaCl. PAP was then added to the concentrated protein for a final concentration of 4 mM.
The SULT2B1b fusion protein was expressed, and cells were disrupted using a very similar protocol but with a lysis buffer containing 50 mM Tris, pH 7.5, 500 mM NaCl. Soluble protein was loaded onto glutathione-Sepharose 4B resin and extensively washed in lysis buffer and 0.5 mM PAP. The pure SULT2B1b enzyme was then eluted by thrombin digestion. PAP was added to the eluted protein for a final concentration of 1 mM to increase the protein solubility. The protein then was concentrated to ϳ15 mg/ml and run over a Superdex 75 HR 10/30 column with 25 mM Tris, pH 7.5, and 100 mM NaCl. The fractions corresponding to SULT2B1b were concentrated to 23 mg/ml, and PAP was added for a final concentration of 4 mM.
Protein Crystallization-Crystals of SULT2B1a were grown by the hanging drop vapor diffusion method by mixing 2 l of prepared SULT2B1a with 2 l of reservoir solution containing 0.8 -1.0 M sodium tartrate, 0.2 M Li 2 SO 4 , and 0.1 M CHES, pH 9.0. Crystals were transferred in four steps of increasing sodium tartrate and ethylene glycol into 1.5 M sodium tartrate, 15% ethylene glycol, 40 mM Li 2 SO 4 , 4 mM PAP, and 80 mM CHES, pH 9.0, and then flash-cooled in a stream of nitrogen gas cooled to Ϫ180°C for data collection. Attempts to cocrystallize full-length SULT2B1a in the presence of pregnenolone resulted in crystals, but no substrate was observed in the acceptor substrate binding pocket. In addition, attempts to soak pregnenolone into the crystal also failed. However, it was discovered that a molecule of CHES from the buffer was binding with the sulfate group interacting with Gln 170 and Tyr 29 in a competitive manner to the acceptor substrate in the binding pocket. Attempts to co-crystallize the complex in Tris buffer did not result in diffraction quality crystals.
Crystals of SULT2B1b were obtained using the hanging drop method by mixing 2 l of prepared protein with 2 l of reservoir solution containing between 0.55 and 0.8 M sodium citrate and 0.1 M imidazole where I i is the intensity of the ith observation and ͗I͘ is the mean intensity of the reflection. b R cryst ϭ ͚F o |Ϫ|F c /͚|F o | calculated from working data set. R free is calculated from 5% of data randomly chosen not to be included in refinement.
at pH 8.0 -8.5. For co-crystallization attempts, 1 mM pregnenolone, 1 mM DHEA, saturated cholesterol (Ͻ1 mM), and saturated 25-hydroxycholesterol (Ͻ1 mM) were added to the protein solution of different hanging drops. These crystals were transferred in four steps of increasing sodium citrate and ethylene glycol concentrations into the cryosolution consisting of 1.0 M sodium citrate, 0.1 M imidazole, pH 8.0, 50 mM NaCl, 1 mM PAP between 1 mM and saturated acceptor substrate, and 10% ethylene glycol. Crystals were flash-cooled directly into liquid nitrogen and then transferred into a stream of nitrogen gas cooled to Ϫ180°C. Electron density was visible for the pregnenolone and DHEA data sets. The 25-hydroxycholesterol data set did not produce electron density for the substrate; however, it diffracted to higher resolution than any apocrystals tried, therefore it is used as the apodata set.
Data Collection and Refinement-The SULT2B1b data set collected in the presence of pregnenolone was collected on a MARCCD area detector system at the Southeast Regional Collaborative Access Team beamline 22 at Advanced Photon Service. All of the other data sets were collected on a Rigaku RU3H rotating anode generator with a Raxis4 area detector and MSC blue mirrors ( Table I). All of the diffraction data were processed using HKL2000 (17). The phase problem for the SULT2B1a data set was solved by molecular replacement using the program AMoRe (18) from the CCP4 package (19) with the x-ray crystal structure of SULT2A1 solved in the presence of PAP and with no acceptor substrate as the starting model (20). This model was refined at 2.9-Å resolution using iterative cycles of model building with the program O (21), and torsion angle simulated annealing, energy minimization, and individual B-factor refinement using the program CNS (22). The model of SULT2B1a was then used to refine the structure of SULT2B1b in the presence of pregnenolone at 2.3 Å in a similar fashion. This structure was used as the starting model to solve the structures of the apo and DHEA data sets of SULT2B1b. The quality of the models was checked using PROCHECK (23). Ordered residues for the SULT2B1a structure are Leu 12 -Leu 95 and Ser 103 -Glu 296 . Ordered residues for the SULT2B1b structure with pregnenolone bound are Ser 18 (from the thrombin cleavage site) and Asp 19 -Glu 311 . Ordered residues for the SULT2B1b structure with DHEA are Ser 18 , Asp 19 -Asp 115 , and Ser 118 -Glu 311 . The ordered residues for the SULT2B1b structure with no acceptor substrate present are Leu 27 -Asp 115 and Pro 119 -Glu 311 .
Sulfotransferase Assay-Sulfotransferase activity of truncated SULT2B1b was determined using radiolabeled cholesterol and pregnenolone according to a previously reported procedure (5). 20-l reaction volumes contained 0.4 g/tube SULT2B1b and either cholesterol or pregnenolone at 5 M. Reactions were carried out at 37°C for 5 min. This construct was shown to have two-thirds the activity for cholesterol and four times the activity for pregnenolone as the full-length protein (Table II). Activity for the full-length SULT2B1a used in these experiments has been reported previously (5).

RESULTS
Overall Description-The crystal structures of full-length SULT2B1a and the construct Asp 19 -Asp 312 of SULT2B1b with PAP and no acceptor substrate bound are for all intents and purposes identical. For SULT2B1a, the amino-terminal 11 residues and the carboxyl-terminal 52 residues (proline-rich region) are disordered. For SULT2B1b, residues Asp 13 -Lys 26 are disordered. Therefore, the sequences of the ordered residues are identical. Because the structure of SULT2B1b is a higher resolution structure, the discussion of the structures will focus on the SULT2B1b structures.
The overall structure of SULT2B1b is that of the classical cytosolic sulfotransferase fold consisting of an ␣/␤ motif comprised of a central 5-stranded parallel ␤-sheet (Fig. 1). Superposition to other cytosolic sulfotransferases reveals a r.m.s. of 1.3 Å for 272 structurally equivalent C␣s to human estrogen sulfotransferase (EST) (24) and 1.5 Å for 263 structurally equivalent C␣s to SULT2A1 (25). As might be expected, residues surrounding the PAPS binding site are highly conserved among the cytosolic sulfotransferases with the major structural and sequence differences existing around the hydrophobic acceptor substrate binding pocket.
PAP Binding-The loop connecting the first strand of the central ␤-sheet with the first helix forms the PSB-loop (residues 67 TYPKSGT 73 ) (26). This loop is similar in structure to the P-loop motifs found in many protein kinases and is involved in binding the 5Ј-phosphate of PAPS (or PAP) (27). A number of residues from the PSB-loop form interactions with the 5Ј-phosphate of PAP (Fig. 2a). Backbone amide nitrogens from Ser 71 (3.1 Å), Gly 72 (3.1 Å), and Thr 73 (2.8 Å) as well as the hydroxyl oxygen of Thr 73 (2.5 Å) are all within hydrogen-bonding distance to the oxygen atom O4P of the 5Ј-phosphate. The O5P oxygen of the 5Ј-phosphate is in position to form hydrogen bonds with both the backbone amide nitrogen (2.9 Å) and the hydroxyl oxygen (2.8 Å) of Thr 74 . The amine nitrogen from highly conserved Lys 70 is within hydrogen-bonding distance (3.0 Å) to the oxygen atom O6P of the 5Ј-phosphate. From a previously solved structure of human EST in complex with PAPS, it has been shown that this oxygen is the bridging oxygen between the 5Ј-phosphate and the sulfuryl moiety ( Fig.  2b) (24).
A number of residues are also involved in binding the 3Јphosphate. Backbone amide nitrogens from Lys 275 (2.8 Å) and Gly 276 (2.9 Å) are in position to bind the O2P phosphate oxygen, whereas the O3P phosphate oxygen is within hydrogen-bonding distances to nitrogen atom NH 2 of Arg 274 (3.1 Å) and atom NH 2 of Arg 147 (3.1 Å). The O1P oxygen of the 3Ј-phosphate is 3.1 Å from nitrogen atom NE of Arg 274 and is also 2.8 Å from the hydroxyl oxygen of the highly conserved residue Ser 155 .
The adenine group of PAP is found sandwiched between two aromatic residues in a parallel ring-stacking orientation with Trp 75 and anti-parallel ring-stacking orientation with Phe 246 .

FIG. 1. Ribbon diagram of SULT2B1b co-crystallized in the presence of substrate donor product PAP (orange) and the acceptor substrate pregnenolone (light blue).
The amino-terminal helix (red) is disordered in the SULT2B1b ϩ PAP complex but becomes ordered upon acceptor substrate binding. Catalytic residues Lys 70 and His 125 are also pictured (cadetblue). This figure was created using MOLSCRIPT and Raster3D (31,32). Atom N3 of the adenine is hydrogen bound to the phenolic hydroxyl of Tyr 210 (2.8 Å), and the N6 amine nitrogen is 2.9 Å from the carbonyl oxygen of Ser 244 . Substrate Binding-The crystal structures of SULT2B1b in the presence of PAP and acceptor substrates pregnenolone or DHEA reveals that these two similar compounds, which differ at the C17 carbon side chain, bind in identical positions and orientations (Fig. 3, a-c). The substrate binding site is highly hydrophobic. There is only one hydrogen bond (2.7 Å) made between the protein and substrate that is between the O3 acceptor hydroxyl off the C3 carbon and atom NE2 of His 125 , the proposed catalytic base. Residues found lining the hydrophobic acceptor binding pocket near the A ring of the steroid substrates other than His 125 are Phe 272 , Tyr 159 , Gln 165 , Tyr 44 , and Trp 103 . The aromatic side chain of Trp 103 is stacked parallel to that of the steroids A and B rings (Fig. 3a). Residues found lining the pocket near the B and C rings are Tyr 257 , Leu 260 , and Thr 106 . Finally, residues found lining the opening of the binding pocket at the surface of the protein near the D ring of the steroid are Trp 98 , Val 108 , Leu 43 , Leu 264 , and Ile 20 . Interestingly, Ile 20 resides on the amino-terminal helix comprised of residues Asp 19 to Lys 26 , which is only ordered upon substrate binding and is required for cholesterol activity. In both the SULT2B1a and SULT2B1b structures with no acceptor substrate, the first ordered residue is the leucine equivalent to that of Leu 27 in SULT2B1b.

DISCUSSION
Catalytic Mechanism-Based on the crystal structure of mouse EST with PAP and estradiol-17␤ (E2) bound, it has previously been suggested that sulfotransferases proceed by an S N 2-like in-line displacement reaction mechanism (27). The position of pregnenolone binding in SULT2B1b places the acceptor hydroxyl 4.3 Å away from the position of the bridging oxygen between the 5Ј-phosphate and the sulfuryl group in PAPS. Superposition of PAPS from the human EST ϩ PAPS structure on to the SULT2B1b active site positions the sulfur atom of PAPS 2.6 Å from the acceptor hydroxyl in line with the acceptor hydroxyl and the bridging oxygen from the leaving group PAP (Fig. 2b) (24). Thus, this geometry supports the in-line displacement mechanism.
In such a reaction mechanism for SULT2B1b, conserved His 125 could serve as a general base to help deprotonate the acceptor hydroxyl and conserved Lys 70 could help to reduce the negative charge build up on the bridging oxygen between the leaving group PAP and the sulfuryl group of PAPS. Interestingly, in the structure of PAPS bound to human EST, residues equivalent to conserved Ser 155 , which interacts with the 3Ј-phosphate of PAP, and Lys 70 form a hydrogen bond (24). However, in the transition state and PAP bound states, the lysine side chain undergoes a conformational change and interacts with the bridging oxygen between the 5Ј-phosphate and the sulfuryl group to help facilitate sulfuryl transfer (24, 28). Thus, residue Ser 155 may participate not only in substrate donor binding but may help inhibit hydrolysis of PAPS by interacting with the lysine in the absence of acceptor substrate (Fig. 2b).
Comparision to Human EST and SULT2A1-Although the overall fold of the SULT2B1 isoforms is very similar to that of other cytosolic sulfotransferases, there is a striking difference in the position of the acceptor substrates (Fig. 4). Three other cytosolic sulfotransferases that have been solved in the presence of their presumed physiological substrates are human EST and mouse EST in the presence of E2 and human SULT2A1 in the presence of DHEA (24,25,27). Interestingly, DHEA is found in two different competitive orientations in the active site of SULT2A1 (25). One site positions the acceptor hydroxyl in the same position as that of E2 in the human EST structure, and the molecule extends out toward the surface of the protein in the same direction and position as E2 but rotated by approximately 30°with respect to an imaginary axis running the length of the molecules. This is believed to be the catalytically relevant binding orientation of DHEA (25). The other position of the DHEA molecule binds such that the C3 carbon only differs in position by 1.1 Å; however, the molecule has shifted such that the O17 ketone oxygen has moved 7.4 Å from that of the catalytic binding position resulting in an approximate 45°angle difference in binding. This positions the acceptor hydroxyl 2.2 Å away from the catalytic position, no The hydrogen bond between the catalytic histidine 125 and the O3 oxygen of the pregnenolone is shown by a gray line. b and c, pictured is the electron density from simulated annealing F o Ϫ F c omit maps contoured at 3 from the refinement of the data set with pregnenolone bound (b) and DHEA bound (c). This figure was created using MOLSCRIPT and Raster3D (31,32). longer in line with the position of the leaving group PAP of PAPS. Thus, this position has been suggested to be involved in substrate inhibition. Interestingly, the catalytic binding position of pregnenolone and DHEA in the SULT2B1b structures is in the same position but a different orientation as the substrate inhibition site of SULT2A1 (Fig. 4). However, in the case of SULT2B1b, the acceptor hydroxyl is in the correct position and orientation for catalysis (Fig. 2b). In addition, the pregnenolone and DHEA are rotated approximately 170°along the imaginary long axis of the molecule with respect to DHEA binding in the substrate inhibition site in SULT2A1.
Structural difference at the opening of the binding pocket suggest that residues Gln 165 , Leu 43 , Leu 27 , and Leu 264 of SULT2B1b would form steric clashes with the D ring of a steroid if it were bound in the catalytic orientation as E2 in human EST and DHEA in SULT2A1 (Fig. 4). Perhaps the most striking structural difference in the overall fold in this protein is the position of the amino-terminal helix of SULT2B1b with respect to human EST and SULT2A1. Structural similarities between the proteins begin at residue Glu 30 of SULT2B1b. In the acceptor substrate bound structure of SULT2B1b, the amino-terminal helix runs across the opening of the acceptor substrate binding pocket positioning itself between loops containing residues Leu 259 to Ser 263 and Thr 106 to Val 108 , effectively burying the substrate in the hydrophobic binding site (Fig. 3a).
Substrate Specificity of SULT2B1a and SULT2B1b-Previously, it has been suggested that the specificity difference for cholesterol and pregnenolone between SULT2B1a and SULT2B1b could be traced to the unique amino-terminal residues 19 DISEI 23 (5). The results from alanine-scanning mutagenesis of the 19 DISEI 23 region reveal that only the I20A and I23A mutants knocked out cholesterol-sulfonating activity. It was then shown that this activity could be partially restored by replacement with a conservative substitution such as leucine (5). The position of Ile 20 and Ile 23 are such that they lie on the same side of the helix facing on the inside of the hydrophobic pocket, whereas residues Ser 21 and Glu 22 are solvent-exposed. Interestingly, the residues corresponding to 19 DISEI 23 of SULT2B1b are 4 PPPFH 8 in SULT2B1a. With three prolines in a row, it is unlikely that these residues are able to form a ␣-helix and therefore would be unable to cover the opening to the substrate binding pocket in the same manner as is seen in the SULT2B1b structure. Although we do not have a crystal structure of SULT2B1b with cholesterol bound yet, the position of pregnenolone is such that the O20 ketone is only 3.4 Å from atom CD1 of Ile 20 and the C21 atom is 6.0 Å from atom CD1 of Ile 23 . Thus, these residues may be in position to form positive van der Waal's interactions with the longer side chain of the C17 atom of cholesterol. However, slight conformational changes of the protein may be necessary to accommodate the longer side chain of cholesterol.
Alternatively, it is possible that Ile 20 may form positive van der Waal's interactions with the hydrophobic side chain of cholesterol, whereas the role of Ile 23 is to form positive van der Waal's interactions with residues Leu 27 , Gly 42 , and Leu 43 locking the helix in place. Neither interaction alone may be enough to bind the helix in the present orientation upon substrate binding, but when both are present the helix can lock down in position over the active site to bind the hydrophobic cholesterol molecule. To better understand the possible binding position of the cholesterol side chain, coordinates of cholesterol were taken from the structures of cholesterol bound to ␤-cryptogein (29) and bound to nuclear receptor retinoic acid-related orphan receptor ␣ ligand binding domain (30) and superimposed on the pregnenolone molecule bound to SULT2B1b. In these superpositions, the closest distance between Ile 23 and cholesterol is 5.9 Å with the side chains of the cholesterol molecules extending between residues Trp 98 and Val 108 toward the surface of the protein (Fig. 5). This finding supports the notion that Ile 23 interactions with the protein may help stabilize Ile 20 in an orientation where it can form positive van der Waal contacts with the cholesterol molecule and that Ile 23 does not interact directly with the cholesterol molecule. However, it should be kept in mind that the cholesterol side chain is flexible and conformations other than what are seen in Fig. 5 could be possible.
In conclusion, these structures reveal a different catalytic binding orientation of the acceptor substrate than previously seen and provide insight into the role the amino terminus plays in dictating substrate specificity for cholesterol between the SULT2B1a and SULT2B1b proteins. In addition, these structures combined with previous structures of other cytosolic sulfotransferases help us to understand better the broad overlapping substrate specificities of these enzymes.