Human cytochrome P450 3A7 binding four copies of its native substrate dehydroepiandrosterone 3-sulfate

In addition to the N-terminal truncation typically employed to facilitate human membrane P450 crystallization, crystallization of CYP3A7 ( 31 ) has in addition required the inclusion of three surface mutations (R69G, C77G, and K244E) to corresponding residues found in CYP3A4 (which does crystallize) and three mutations of long, flexible, high-entropy lysine residues on the proximal surface near where the NADPH–cytochrome P450 reductase (CPR) binds (K421A, K422A, and K424A). In order to assess the functional impacts of these internal mutations, ligand binding affinities were compared for this hexamutant and the parental construct designated as the wildtype enzyme. As often observed when substrates bind in the active site and displace the water from the heme iron, addition of the endogenous substrate DHEA-S shifted the Soret peak to shorter wavelengths for both wildtype and hexamutant CYP3A7. Typically monitored as difference spectra collected during titration with ligand ( Fig. S1 ), the peaks at 387 to 388 nm and troughs at ∼420 nm observed with DHEA-S addition are consistent with such a type I binding mode. Plotting the absorbance changes versus DHEA-S concentration ( Fig. 1 A ) revealed that the hexamutant demonstrated a 30% larger absorbance shift (ΔA max ) and an approximately twofold higher affinity ( K d ) than the wildtype enzyme ( Fig. 1 E ). Thus, the endogenous ligand DHEA-S binds to both wildtype and hexamutant forms of CYP3A7, but with slightly different affinities. Figure 1 Ligand binding and DHEA-S catalysis for CYP3A7 wildtype ( red ) and hexamutant ( blue ). A , DHEA-S type I binding is conserved, but the hexamutant is more saturated and has an approximately twofold higher affinity. B , clotrimazole type II binding is conserved and binds to a similar extent to both enzymes but has approximately twofold higher affinity for the hexamutant. Kinetic parameters of DHEA-S 16α-hydroxylation by ( C ) wildtype and ( D ) hexamutant CYP3A7 best fit to the Michaelis–Menten equation with coefficient of determinations ( R 2 ) of 0.979 and 0.938, respectively. E , values for steady-state binding and catalytic parameters with 95% confidence intervals in parenthesis . All experiments were performed in triplicate. CYP3A7, cytochrome P450 3A7; DHEA-S, dehydroepiandrosterone 3-sulfate. Ligand binding and DHEA-S catalysis for CYP3A7 wildtype ( red ) and hexamutant ( blue ). A , DHEA-S type I binding is conserved, but the hexamutant is more saturated and has an approximately twofold higher affinity. B , clotrimazole type II binding is conserved and binds to a similar extent to both enzymes but has approximately twofold higher affinity for the hexamutant. Kinetic parameters of DHEA-S 16α-hydroxylation by ( C ) wildtype and ( D ) hexamutant CYP3A7 best fit to the Michaelis–Menten equation with coefficient of determinations ( R 2 ) of 0.979 and 0.938, respectively. E , values for steady-state binding and catalytic parameters with 95% confidence intervals in parenthesis . All experiments were performed in triplicate. CYP3A7, cytochrome P450 3A7; DHEA-S, dehydroepiandrosterone 3-sulfate. Both wildtype and hexamutant were also evaluated upon addition of the nonselective inhibitor clotrimazole ( 29 ). Soret peak shifts to longer wavelengths are typically observed when the lone-pair electrons from a ligand nitrogen directly coordinate to the heme iron, replacing the water. Such type II binding was observed upon clotrimazole addition to both the wildtype and hexamutant CYP3A7. The difference spectra peaks at 432 to 434 nm and the troughs at 412 nm ( Fig. S1 , C and D ) were consistent with the imidazole nitrogen heteroatom of clotrimazole coordinating the heme iron in both wildtype and the hexamutant. Analysis of these spectral shifts ( Fig. 1 B ) indicated that the extent of clotrimazole binding (ΔA max ) was very similar for CYP3A7 wildtype and hexamutant, but the dissociation constant ( K d ) was again approximately twofold lower for the hexamutant ( Fig. 1 E ). The observation that both the DHEA-S substrate and an azole-containing inhibitor cause spectral changes with the recombinant CYP3A7 enzyme is important in that this was not observed in a previous report ( 31 ). However, those former studies were accomplished with 1.5 μM CYP3A7 in a buffer containing 1 mM DTT, which was later discovered to indirectly interact with the heme iron via an intervening water molecule. This type of interaction would mean that DTT binding is spectrally indistinguishable from the unliganded water-coordinated form and the DTT could have largely outcompeted binding of ligands with lower affinities. The current results, obtained in buffer without DTT, confirmed that both the wildtype and the hexamutant bind the DHEA-S endogenous substrate and the nonspecific inhibitor clotrimazole, although the hexamutant showed slightly higher affinity for both ligands. In order to further evaluate functional differences that might occur because of the six internal mutations required to crystallize CYP3A7, DHEA-S metabolism to 16α-hydroxy-DHEA-S was also evaluated for both proteins. The wildtype and the hexamutant CYP3A7 similarly converted DHEA-S to a single product ( Fig. S2 ) corresponding to the authentic 16α-hydroxy-DHEA-S standard, which was quantified with calibration standards prepared in matrix. This product was not observed when the cofactor NADPH was omitted ( Fig. S2 B ). Product formation rate was plotted as a function of DHEA-S concentration ( Fig. 1 , C and D ), and data were fitted to both the Michaelis–Menten (hyperbolic) and Hill (sigmoidal) kinetic models to obtain the kinetic parameters K m and V max ( Table S1 ). The models were compared using the Akaike information criterion (AIC), and difference of the second-order Akaike information criterion supported the simplest (Michaelis–Menten) model ( Table S1 ). According to the Michaelis–Menten model, wildtype CYP3A7 has a K m of 77.2 μM and a V max of 55.0 pmol/min/nmol P450, whereas the hexamutant has a K m of 214 μM and a V max of 4.23 pmol/min/nmol P450 ( Fig. 1 E ). This suggests that the six internal mutations were detrimental to product formation, reducing the V max by ∼13-fold and increasing the K m by ∼2.8-fold. Human CYP3A7 modified by truncation of the N-terminal transmembrane helix and addition of a His tag at the C terminus like almost all other crystallized human P450 enzymes is recombinantly expressed in good yields and highly purified with good stability but has been resistant to crystallization despite significant efforts by multiple laboratories. Only recently was the first CYP3A7 structure reported after incorporating six amino acid mutations in the catalytic domain ( 31 ). This structure did not include relevant substrates or inhibitors, but the CYP3A7 active site did contain an adventitious DTT molecule. Herein, the same hexamutant CYP3A7 protein described ( 31 ) was generated and cocrystallized with its native substrate DHEA-S, yielding suitably sized rhombic plates. X-ray data were collected from a single crystal, solved by molecular replacement, and refined ( Table 1 ) to yield a 2.6 Å structure. Table 1 X-ray data collection, refinement, and validation statistics Parameter Hexamutant CYP3A7–DHEA-S Data collection  Space group P 2 1  Cell dimensions   a , b , c (Å) 114.52, 219.38, 130.45   α , β , γ (°) 90, 102.17, 90  Resolution (Å) a 39.94–2.60 (2.74–2.60)  Redundancy a 7 (6.9)   R pim a 0.046 (0.878)  Mn (I/SD) a 10.3 (1.0)  CC ½ a 0.999 (0.473)  Completeness a (%) 99.1 (98.4)  Total reflections 1,342,003  Unique reflections 190,994 Refinement  Resolution (Å) 39.94–2.60 (2.74–2.60)  No. of reflections 1,342,003   R work / R free 0.21/0.27  Molecules per asymmetric unit 12  Number of nonhydrogen atoms/ B -factor  Protein 45,416/78  Ligand 600/81  Heme 588/71  Water 32/43  rmsds  Bond lengths (Å) 0.011  Bond angles (°) 1.2  Coordinate error (maximum likelihood) (Å) 0.50  Ramachandran plot: preferred/allowed/outliers (%) 92.03/7.46/0.51 a Highest resolution shell is shown in parenthesis . X-ray data collection, refinement, and validation statistics Highest resolution shell is shown in parenthesis . As observed previously ( 31 ), the CYP3A7 hexamutant adopted a typical P450 fold with the three CYP3A4-mimicking mutations in the β1 loop region (R69G and C77G) and G helix (K244E) and the reductase-binding region mutations (K421A, K422A, and K424A) on the K–L loop region ( Fig. 2 A ). Of the 12 copies of the hexamutant CYP3A7 protein in the asymmetric unit, two distinct types of CYP3A7 molecules were observed. Six molecules (molecules A, B, F, G, K, and L) did not have any density present for a ligand ( Fig. 2 A ). The remaining six molecules (molecules C, D, E, H, I, and J) each had four DHEA-S molecules, which were readily modeled with 100% occupancy ( Fig. 2 B ). Two DHEA-S molecules were in the active site proper, one closer to the heme (DHEA-S1) than the other (DHEA-S2), one in a channel between the active site and the surface (DHEA-S3), and one on the protein surface (DHEA-S4) ( Fig. 2 B ). The arrangement of 12 molecules and crystal packing were shown in Fig. S3 . Seven distinct types of crystal contacts were observed in this assembly. The first two types of interactions involve the dimerization of two liganded molecules. One interface is located on the β1–β2 strands of the two molecules, where a salt bridge is formed through the K390 and D380 side chains ( Fig. S4 A ). The other interface occurs on the D helix side of the protein. Here, the side chain of N341 from the J–K loop of one molecule forms a hydrogen bond with the P147 ketone backbone of the D helix on the other molecule ( Fig. S4 B ). Some additional crystal contacts are present, but they are not well defined because of the flexibility of the side chains. Two additional interactions involve the dimerization of two unliganded molecules. One interaction is facilitated by hydrophobic contacts between the two parallel G' helices of the two molecules ( Fig. S4 C ). Another interaction interface is located at the D helices and J–K loops, both perpendicular to each other, within the two molecules. In this scenario, the N341 side chain of the J–K loop forms hydrogen bonds with the N341 and K342 backbone ketones ( Fig. S4 D ). As before, some additional interactions are not well defined becauseof the flexibility of the side chains on the protein surface. The final three interactions occur between the liganded and unliganded molecules. The first type of interaction transpires at the interface of the F and F' helices in both molecules, where they form hydrophobic interactions ( Fig. S4 E ). The second interaction takes place at the interface of the two perpendicular A helices of the molecules. Here, the E63, K66, and K67 side chains of the unliganded molecule bond with the backbone amide and ketone groups of the β1, A helix, and A’’ helix, respectively, in the liganded molecule ( Fig. S4 F ). The last interaction is observed between the B helix of the liganded molecule and the D helix of the unliganded molecule, where the D88 side chain of the liganded molecule forms a salt bridge with the K143 side chain of the unliganded molecule ( Fig. S4 G ). Figure 2 Hexamutant CYP3A7 structural overview. A , representative ligand-free molecule with the positions of the K421A/K422A/K424A/R69G/C77G/K244E mutations shown in red sticks . B , comparison of a ligand-free molecule ( yellow ribbons ) with a molecule ( green ribbons ) containing four copies of DHEA-S (shown in sticks with gray carbons ). Heme in black sticks with rust sphere for iron. CYP3A7, cytochrome P450 3A7. Hexamutant CYP3A7 structural overview. A , representative ligand-free molecule with the positions of the K421A/K422A/K424A/R69G/C77G/K244E mutations shown in red sticks . B , comparison of a ligand-free molecule ( yellow ribbons ) with a molecule ( green ribbons ) containing four copies of DHEA-S (shown in sticks with gray carbons ). Heme in black sticks with rust sphere for iron. CYP3A7, cytochrome P450 3A7. Pairwise comparisons of CYP3A7 molecules containing DHEA-S ligands ( Fig. 2 B , green ) and those without ligand ( Fig. 2 B , yellow ) consistently yielded Cα rmsd of 0.66 to 0.85 Å, whereas comparisons between ligand-free molecules had rmsd values of 0.35 to 0.6 ( Table S2 ) and comparisons between molecules containing DHEA-S had rmsd values of 0.26 to 0.39 ( Table S2 ). The most significant differences between the liganded and unliganded structures were observed in helices A, A′, B, F′, G′, and G and β 1– 2 ( Fig. 2 B ). The F′, G′, and G helices shift slightly in the opposite direction from the A, A′, B, and β 1–2 secondary structure elements, slightly opening up the channel at the level of DHEA-S3. However, large ligand-triggered structural changes in the backbone were not observed. The DHEA-S ligands are all well defined by the electron density ( Fig. 3 A ), but each has distinct interactions with the protein. Closest to the heme, DHEA-S1 ( Fig. 3 B ) is located 4.1 Å above the heme Fe. Its C17 ketone is directed toward heme iron and the C3 sulfate projecting toward the β4 region. The backbone amide of β4 G481 is positioned to hydrogen bond (2.5–3.1 Å) with either of the two oxygen atoms of the DHEA-S1 sulfate ( Fig. 3 C ). The distance from the site of metabolism, C16, to the heme iron is 5.8 Å ( Fig. 3 B ). The rest of the interactions with DHEA-S1 are largely hydrophobic. Residues within 4 Å of DHEA-S1 include Y57, R105, S119, N214, P215, L216, A305, T309, V369, A370, G480, G481, and L482 ( Fig. 3 C ). DHEA-S2 is located above DHEA-S1 in the active site ( Fig. 3 B ). The distance from its C17 ketone to the heme iron is 8.7 Å ( Fig. 3 B ). The long axis of its steroid nucleus is parallel with the F helix. One side of its steroidal core packs against P215 in the loop between F and F′, whereas the side chain of F108 in the loop between B and B′ packs against the opposite side ( Fig. 3 C ). One oxygen of the DHEA-S2 sulfate hydrogen bonds with the side chain amino group of K224 (2.6–3.2 Å, Fig. 3 C ). Only one of the CYP3A7 molecules (molecule D) containing DHEA-S suggests that another oxygen atom of this sulfate can form a hydrogen bond with the side chain guanidinium group of R106 (2.65 Å, Fig. 3 C ). Residues within 4 Å of DHEA-S2 include D76, R105, R106, F108, S119, I120, F213, P215, V220, K224, F241, I301, F304, and E374 ( Fig. 3 C ). DHEA-S3 lies in the access channel close to the protein surface between the A′ and F′ helices with the ketone at C17 directing toward the F′ helix ( Fig. 3 D ). The C3 sulfate extends to the surface where one of its oxygen atoms is in position to hydrogen bond (∼2.6 Å) with the side chain guanidinium group of R243 in an adjacent molecule ( Fig. 3 C ). Residues within 4 Å of DHEA-S3 include F46, L47, A50, L51, and R54 from helix A, D76 from the β1 system, amino acids L221, K224, and V225 from helix F′, and residues N237, I238, and R243 from the adjacent molecule ( Fig. 3 C ). Finally, DHEA-S4 is located above the protein surface next to helices A and F′ at the interface between adjacent or symmetry-related molecules ( Fig. 3 C ). One sulfate oxygen atom of DHEA-S4 forms a hydrogen bond with the backbone amide of G26 in the adjacent or symmetry-related protein molecule, but the majority of the interactions appear to be hydrophobic. Residues within 4 Å of DHEA-S4 include P45, F46, and L47 from helix A′, V225 and F226 from F′, and I232, A235, and L236 from the adjacent protein molecule ( Fig. 3 C ). This hydrophobic surface is thought to be embedded in the endoplasmic reticulum membrane. Figure 3 Interaction s of the DHEA-S substrates ( sticks with gray carbons ) with the CYP3A7 hexamutant ( rainbow ribbons from blue N terminus to red C terminus). A , strong electron density (simulated annealing composite omit at 1σ) defines the positioning of all four molecules of the substrate DHEA-S. B , the DHEA-S closest to the heme (DHEA-S1) is oriented with its C17 carbonyl directed toward the heme iron. The second DHEA-S molecule (DHEA-S2) also binds in the active site cavity ( gray mesh ) but in a different orientation. C , residues within 4 Å of any of the DHEA-S molecules are shown as sticks , with key sulfate–protein interactions indicated by dashed lines . The sulfate of DHEA-S3 extends to the surface and interacts with Arg243 of a symmetry-related molecule ( black sticks ). D , channels from the active site to the protein exterior are indicated by black arrows 1 to 4. DHEA-S, dehydroepiandrosterone 3-sulfate. Interaction s of the DHEA-S substrates ( sticks with gray carbons ) with the CYP3A7 hexamutant ( rainbow ribbons from blue N terminus to red C terminus). A , strong electron density (simulated annealing composite omit at 1σ) defines the positioning of all four molecules of the substrate DHEA-S. B , the DHEA-S closest to the heme (DHEA-S1) is oriented with its C17 carbonyl directed toward the heme iron. The second DHEA-S molecule (DHEA-S2) also binds in the active site cavity ( gray mesh ) but in a different orientation. C , residues within 4 Å of any of the DHEA-S molecules are shown as sticks , with key sulfate–protein interactions indicated by dashed lines . The sulfate of DHEA-S3 extends to the surface and interacts with Arg243 of a symmetry-related molecule ( black sticks ). D , channels from the active site to the protein exterior are indicated by black arrows 1 to 4. DHEA-S, dehydroepiandrosterone 3-sulfate. In addition to the primary channel containing the DHEA-S molecules ( Fig. 3 C , channel 1), three other channels are apparent from the active site to the protein surface ( Fig. 3 D ). One channel exits between the B–B′ loop and B′–C loop (channel 2). Another passes between the F and I helices (channel 3), and the last one between the B–B′ loop and β 1– 1 (channel 4).

The first CYP3A7 structure, that of the same hexamutant used herein, had two nearly identical molecules (Cα rmsd 0.345), both with the buffer component DTT in the active site but not directly interacting with the heme or polypeptide ( 31 ). The crystal packing differs slightly between the two structures. The CYP3A7/DTT structure contains only two protein molecules in the asymmetric unit ( 31 ), whereas our structure hosts 12 molecules. Two types of crystallographic dimers were identified in the DTT structure. One contact area is constituted by the hydrophobic residues from the antiparallel G′-helices, which is also observed in our structure. Another interface is formed by the antiparallel D-helices with a sulfate ion situated in between, linking the two helices via hydrogen bonds and salt bridges. In our structure, we also identified an interface at the D-helix region, but in our case, the two D-helices are oriented perpendicularly to each other. Hydrogen bonds are found either between the D-helix and J–K loop or between both J–K loops in the two molecules. Our structure also presents additional interfaces involving two F–F′ regions and A-helices. The discrepancies in crystal packing between the two structures are likely attributed to the different ligands they cocrystallized with. These ligands possibly induce conformational changes in CYP3A7, altering the homogeneity of the protein during crystallization. Superimposing the DTT structure with representative ligand-free and 4xDHEA-S conformations determined herein results in Cα rmsd values of 0.44 Å and 0.723 Å, respectively. Thus, few structural differences occur between the ligand-free and DTT structures despite different crystallization conditions, different space groups, and different packing interactions in the crystalline lattice. While the C77G, K421A, and K424A mutations did not form intermolecular contacts between molecules in the DTT complex, herein five of the mutated residues were involved in intermolecular contacts, with the exception being C77G. Regardless, the most substantial differences between the DTT and 4xDHEA-S structures occur in helices A′, B′, F′, F, G′, and G, β 1– 1 , β 1– 2 , and their loop regions ( Fig. 4 A ). These helices, β sheets, and loops all slightly move outward to open up the main channel where DHEA-S binds but particularly in the region of DHEA-S3 ( Fig. 4 A ). Figure 4 Comparisons of CYP3A7 hexamutant structures determined with different liganded states. A , the structure with four DHEA-S molecules bound ( green ) is quite similar to the previous structure determined with the adentitious ligand DTT ( slate ). B , comparison of the previous DTT (154 g/mol) complex ( slate ), the DHEA-S (288 g/mol) complex ( green ), and the ligand-free structure ( yellow ) reveal remarkably little changes in active site residue rotamers (DTT not shown). DHEA-S, dehydroepiandrosterone 3-sulfate; CYP3A7, cytochrome P450 3A7. Comparisons of CYP3A7 hexamutant structures determined with different liganded states. A , the structure with four DHEA-S molecules bound ( green ) is quite similar to the previous structure determined with the adentitious ligand DTT ( slate ). B , comparison of the previous DTT (154 g/mol) complex ( slate ), the DHEA-S (288 g/mol) complex ( green ), and the ligand-free structure ( yellow ) reveal remarkably little changes in active site residue rotamers (DTT not shown). DHEA-S, dehydroepiandrosterone 3-sulfate; CYP3A7, cytochrome P450 3A7. One feature reported in the previous DTT structure was that F304 in the CYP3A7 I helix flanking the active site adopts a rotamer positioning the aromatic ring toward the F’/G′ active site “roof” where it interacts with two leucines and two other phenylalanines from those structural elements. This was notable because F304 adopts different rotamers in CYP3A4 and CYP3A5. Depending on the size of the active site ligand, different F304 rotamers are observed, and it is thought to be a gatekeeper for access of larger ligands into the active site proper. However, comparison of F304 in all three CYP3A7 structures now available shows that F304 maintains its upward rotamer forming part of this hydrophobic patch ( Fig. 4 B ) including residues F189, F203, L210, L211, F213, L249, V253, F241, L295, and M296. This feature of CYP3A7 and related interactions have been suggested to confer greater rigidity to CYP3A7 ( 31 ) associated with decreased substrate metabolism compared with CYP3A4. The structure herein definitively demonstrates that the hexamutant CYP3A7 binds four copies of DHEA-S simultaneously, two in the active site, one in a channel to the surface, and one at the membrane-binding surface. A relevant question is whether this is also the case for the wildtype CYP3A7 in the absence of the six internal mutations. DHEA-S1 is buried in the active site cavity adjacent to the heme and surrounded by highly hydrophobic side-chain residues from helices A, F, G, and I and β1. This active site is exactly where one would expect to find substrate binding, and DHEA-S1 is in addition oriented with the C16 site of observed metabolism directed toward the heme iron as would be required for 16α-hydroxylation. This occurs even though there are not notable specific interactions between DHEA-S1 and side chains lining the active site, only the backbone of the β4 loop. Since DHEA-S1 is closest to the heme iron, this ligand is most likely primarily responsible for the spin change observed as a spectral shift in the DHEA-S binding studies. However, even though the nearest mutation K244E is 25.9 Å away from DHEA-S1, clearly one or more of the ligands indirectly facilitate this spin change because the affinity and saturation as measured by spectral shift are somewhat greater in the hexamutant than in the wildtype ( Fig. 1 A ). Based on proximity and relative orientation, it appears that the presence of DHEA-S2 could facilitate positioning of DHEA-S1 near the iron. If DHEA-S2 were not present, DHEA-S1 would certainly have more freedom to move about large active site cavity; however, no evidence of cooperativity was observed in the binding studies consistent with the multiple ligand active site binding model ( 32 ). This may be because the binding of DHEA-S1 and DHEA-S2 is totally independent and binding of one (likely DHEA-S2) is spectroscopically silent. There are other examples of cytochrome P450 enzymes with more than one ligand in the active site, including steroids or synthetic sterol dimers ( 33 , 34 ). While this would not explain the lack of observed cooperativity in the DHEA-S oxidation studies, such catalysis is further complicated by the involvement of CPR and cytochrome b 5 , making a direct comparison of affinities and stoichiometries difficult. Binding of DHEA-S3 in the channel and DHEA-S4 on the surface was somewhat unexpected and also likely to be spectroscopically silent. The binding site for DHEA-S3 is closest to one of the mutations. DHEA-S3 is located directly above the C77G mutation site on β1 loop region ( Fig. 2 A and 2 B ). When the six mutations were modeled back to the corresponding wildtype residues, most were easily accommodated. However, regardless of the rotamer modeled, a clash was observed between the C77 side chain and DHEA-S3. Thus, it is possible that either the wildtype enzyme might be incompatible with the binding of DHEA-S3 in the access channel or the wildtype might undergo a conformational change opening up the ligand access channel slightly to accommodate DHEA-S3. However, this primary channel is very similar in CYP3A4 and CYP3A5 without these internal mutations ( 31 ). DHEA-S4 is located on the protein surface adjacent to the F′ and G′ helices. The side chains on the outside of these two short helices and the loop in between are highly hydrophobic, which provide a favorable environment for DHEA-S4 to bind. Surrounding the other sides of DHEA-S4 is the same region of three adjacent molecules in the crystalline lattice. Without the convergence of these intermolecular contacts, a large part of DHEA-S4 would be exposed to the hydrophilic solvent and, therefore, the binding of DHEA-S4 on the protein surface might be disrupted. However, in the native environment, this hydrophobic surface of the F′–G′ region is thought to be embedded in the hydrophobic core of the endoplasmic reticulum membrane and likely to be the dynamic region whereby hydrophobic substrates like steroids pass from the membrane into the cytochrome P450 active site. A number of membrane P450 enzymes have now been observed with channels in the same structural region as the one filled with DHEA-S molecules here. An early structure of the 87% identical adult CYP3A4 enzyme also defined the steroid progesterone binding site on its F′–G′ surface (Protein Data Bank [PDB] entry: 1W0F ( 35 )). In that structure, the oxygen of progesterone’s acetyl side chain hydrogen bonded with the amide nitrogen of D214 in the F–F′ loop. This binding area was proposed to serve as an effector or substrate-binding site in the original publication ( 35 ). This idea is consistent with a significant body of functional data indicating that CYP3A4 demonstrates both heterotropic and homotropic cooperativity, often with steroids. Testosterone can bind cooperatively to CYP3A4 ( 36 ). Progesterone ( 37 , 38 , 39 ), testosterone ( 40 ), and 17β-estradiol ( 40 ) metabolism all demonstrates homotropic cooperativity, whereas progesterone heterotropically stimulates estradiol ( 41 ) and carbamazepine ( 42 ) metabolism. Site-directed mutation results also support the importance of this peripheral binding site as mutation of L211 and D214 to Phe and Glu, respectively, affects progesterone cooperativity in both binding and metabolism ( 43 ). Thus, substantial experimental evidence supports progesterone high-affinity binding on the peripheral surface and its role as a CYP3A4 allosteric activator, enhancing CYP3A4 activity toward other substrates ( 44 , 45 ). Much less information is available regarding the phenomenon of cooperativity in CYP3A7 catalysis. Herein, the internal mutations clearly cause a ∼13-fold decrease in the V max and threefold increase in the K m for DHEA-S 16α-hydroxylation. This is not surprising as three of the mutations to promote crystallization (K421A/K422A/K424A) are part of the proximal surface where electrostatic interactions with NADPH–CPR allow it to bind and deliver electrons required for catalysis. When electrons are delivered using cumene hydroperoxide as an alternate, the kinetic constants can be similar to the wildtype enzyme ( 31 ). Despite this reduced activity for the hexamutant, neither it nor the more-active wildtype CYP3A7 demonstrated sigmoidal steady-state kinetics; the data yielded a hyperbolic curve best fit to the Michaelis–Menten equation rather than a cooperative model with a Hill coefficient ( Fig. 1 and Table S1 ). Examination of higher concentrations was limited by substrate solubility and lower concentrations by relatively low turnover, but a previous report similarly reported that DHEA-S metabolism was hyperbolic ( 46 ). However, this same report also discovered that the presence of DHEA-S strongly activated CYP3A7-mediated 10,11-epoxidation of the anticonvulsant carbamazepine (5 H -dibenzo[ b , f ]azepine-5-carboxamide) in a concentration-dependent manner ( 46 ). This report concluded that while DHEA-S and carbamazepine are both substrates, they do not inhibit each other’s metabolism. However, DHEA-S decreases the K m of carbamazepine threefold, indicating that one molecule of each substrate could be present simultaneously in the CYP3A7 active site and that either a conformational change induced by DHEA-S or an interaction between DHEA-S and carbamazepine in the active site could be implicated. Carbamazepine is a fused aromatic planar molecule with a molecular weight of 236 g/mol and is thus similar in size and shape to DHEA-S. Thus, the current structure suggests it could bind similarly to one of the two DHEA-S molecules observed in the active site, which would support the multiple ligand active site model. It is also possible that one of the outer DHEA-S3 or DHEA-S4 molecules indirectly alters carbamazepine metabolism in the active site, analogous to progesterone in CYP3A4. However, the authors of the previous study found that the sulfate of DHEA-S was critical in decreasing the K m for carbamazepine. While DHEA-S3 and DHEA-S4 sulfates do not directly interact with the CYP3A7 monomer, the sulfate of DHEA-S2 makes a direct interaction with Lys224 or R106 in the F′ and B′ helices, respectively. Both F′ and B′ are dynamic regions thought to be involved in sequestering ligands in the active site.

DHEA-S is present in high concentrations in the fetal and neonatal periods, and when xenobiotics including drugs are also present, their CYP3A7 intersection could lead to issues in multiple different ways. First, it is possible that high DHEA-S levels could competitively inhibit CYP3A7-mediated metabolism of drugs; for example, Nakamura et al . ( 46 ) suggested that the effect of endogenous DHEA-S on CYP3A7-mediated drug metabolism may not be uniform. Second, drugs inhibiting CYP3A7 could interfere with DHEA-S metabolism and lead to renal insufficiency in the neonate and developing infant, leading to adverse effects. Finally, environmental xenobiotics activated by CYP3A7 or cleared by CYP3A7 may play important roles in fetal exposure. Therefore, the structure herein is a step forward toward understanding the complex interactions of CYP3A7 with its primary substrate and identifies multiple binding site that may interact with each other and with other substrates bound simultaneously, but numerous CYP3A7–ligand interactions remain to be examined before this information can be translated to improve pediatric pharmacology.

The DHEA-S standard and ampicillin were purchased from Cayman Chemical. The metabolite standard 16α-hydroxy-DHEA-S was obtained from Steraloids. The IPTG, glucose-6-phosphate, and NADP + were acquired from Alfa Aesar. The glucose-6-phosphate dehydrogenase and the internal standard (IS), estriol 3-sulfate, were purchased from MilliporeSigma. The 1,2-dilauroyl- sn -glycerol-3-phosphocholine, 1,2-dioleoyl- sn -glycero-3-phosphocholine, and 1,2-dilauroyl- sn -glycero-3-phosphoserine were obtained from Avanti Polar Lipids. All other chemicals and solvents used were obtained from standard suppliers and were of reagent or analytical grade. The human CYP3A7 complementary DNA (cDNA) was synthesized with codon optimization for Escherichia coli expression (Genscript). Nucleotides coding for wildtype amino acids 1 to 25 composing the transmembrane helix were truncated and replaced with nucleotides coding for the sequence MAKKTSS. Nucleotides were also added at the C terminus to code for a six-residue histidine tag just prior to the stop codon. For the hexamutant, K421A/K422A/K424A/R69G/C77G/K244E mutations were additionally incorporated. Both wildtype and hexamutant genes were cloned into the pCWori+ vector for protein expression. Wildtype CYP3A7 was expressed and purified as described ( 29 ). The hexamutant was expressed and purified as described ( 31 ) with the following modifications to be more like the wildtype protocol. The modifications are (1) DH5α/pGro7 cells were used for expression instead of C41 cells; (2) 4 mg/mL arabinose was added to induce GroEL/ES expression; (3) cells were grown at 25 °C, 190 rpm for 48 h after induction; (4) 80 mM histidine was used in the nickel–nitrilotriacetic acid (Ni–NTA) elution buffer instead of 30 mM; (5) the Ni–NTA column was washed with four column volumes of Ni–NTA lysis buffer containing 0.2% Nonidet P-40 and then five column volumes with the latter buffer supplemented with 5 mM histidine; (6) the CM column wash buffer contains 5 mM potassium phosphate, pH 7.2, instead of 50 mM potassium phosphate, pH 7.4; and 7) the eluted protein from CM column was injected onto a size exclusion column (GE Healthcare HiLoad 16/60 Superdex 200 pg) equilibrated and run with CM elution buffer, yielding highly purified trimeric protein. Full-length human CPR with the K59Q mutation to reduce proteolysis was expressed and purified as described ( 47 ). The full-length human cytochrome b 5 cDNA was synthesized (Genscript) with codon optimization for E. coli expression and with the addition of an N-terminal histidine tag. The b 5 cDNA cloned into the pET15b plasmid was expressed in C41(DE3) E. coli cells. Briefly, an overnight culture (10 mL) was used to inoculate 1 L of Terrific Broth medium containing 200 μg/mL ampicillin and 0.4% glycerol. Culture was grown with shaking at 125 rpm and 24 °C until the absorbance reached 1.0 at 600 nm, then IPTG (0.2 mM) was added to induce protein expression. The induced culture was incubated for another 15 h at 125 rpm and 24 °C. The bacterial cells were harvested by centrifugation at 4000 g for 20 min. The recombinant b 5 protein was then purified from the bacterial cell pellet as described ( 48 ) and using Ni–NTA affinity chromatography. Concentration of the purified b 5 protein was determined using an extinction coefficient of 185 mM −1  cm −1 between the ferrous and ferric forms ( 49 ). Ligand-binding affinities were determined by monitoring shifts in the CYP3A7 heme absorbance spectrum as described previously ( 29 ). DHEA-S (dissolved in dimethyl sulfoxide) was titrated into 1 μM wildtype or hexamutant CYP3A7 in 50 mM potassium phosphate, pH 7.4, 20% glycerol in a 1 cm path length quartz cuvette. Because it binds so tightly, clotrimazole (dissolved in dimethyl sulfoxide) was titrated into 0.1 μM wildtype or hexamutant CYP3A7 in 50 mM potassium phosphate, pH 7.4, 20% glycerol in a 5 cm path length cuvette. Each titration experiment was repeated in triplicate on different days. Changes in absorbance (absorbance maximum minus minimum) were plotted against ligand concentration and fit using the one-site specific binding equation from GraphPad Prism (version 9.3.0 for Mac OS X) (GraphPad Software, Inc) to determine the maximum absorbance change and dissociation constant. The purified recombinant wildtype and hexamutant CYP3A7 enzymes were each reconstituted in a 5× concentrated liposome premix in the presence of the human CPR and cytochrome b 5 following the procedure described by Shaw et al . ( 50 ). A 1 mg/mL liposome stock solution was prepared as 1:1:1 (v/v/v) 1,2-dilauroyl- sn -glycerol-3-phosphocholine, 1,2-dioleoyl- sn -glycero-3-phosphocholine, and 1,2-dilauroyl- sn -glycero-3-phosphoserine. The 5× protein premix was composed of 1 μM recombinant CYP3A7, 4 μM human CPR, and 1 μM human cytochrome b 5 in 0.1 mg/mL liposomes, 3 mM glutathione, 0.5 mg/mL CHAPS, and 20 mM potassium Hepes (pH 7.4). The kinetic reactions (150 μL) were carried out by diluting the 5× protein premix in 100 mM potassium phosphate buffer, pH 7.4, and 3 mM MgCl 2 , and contained a concentration range of DHEA-S (7.5–360 μM). After an equilibration at 37 °C for 3 min, the reactions, prepared in triplicate, were initiated by the addition of an NADPH-regenerating system mix consisting of NADP + , glucose-6-phosphate, and glucose-6-phosphate dehydrogenase (at their respective concentrations in the incubation of 1 mM, 10 mM, and 2 IU/mL). The reactions, done under steady-state kinetic conditions, were incubated for 30 min at 37 °C under agitation and stopped by the addition of ice-cold methanol (200 μL) containing 50 ng/mL of estriol 3-sulfate IS. Incubations without the NADPH-regenerating system mix served as negative controls. Precipitated proteins were collected by centrifugation of the stopped reaction samples for 20 min at 2500 g and 4 °C. Supernatants were transferred to HPLC vials, and aliquots of 5 μL were analyzed by LC–MS/MS. The 16α-hydroxy-DHEA-S metabolite was quantified based on calibration curves ranging from 1 (or 10) nM to 5000 nM. The mean metabolite formation rate obtained from triplicate determinations was fit to the Michaelis–Menten (hyperbolic) and Hill (sigmoidal) equations using GraphPad Prism software (version 9.5.1 for Windows). Statistical analysis for the best fit was done based on the second-order AIC in GraphPad Prism. The DHEA-S incubation samples with the recombinant CYP3A7 enzymes were analyzed by LC–MS/MS with a Waters Acquity ultraperformance liquid chromatography system interfaced by electrospray ionization with a Waters Xevo TQ-S micro tandem quadrupole mass spectrometer (Waters Corp) operating in negative ionization mode and multiple reaction monitoring scan type. The following source conditions were applied 0.5 kV for the capillary voltage, 150 °C for the source temperature, 450 °C for the desolvation temperature, 50 L/h for the cone gas flow, and 900 L/h for the desolvation gas flow. The following mass transitions, collision energies (CEs), and cone voltages (CVs) were used to detect the respective analytes: 383 > 97 (CE = 26 V, CV = 75 V) for the 16α-hydroxy-DHEA-S, 367 > 97 (CE = 34 V, CV = 80 V) for DHEA-S, and 367 > 287 (CE = 28 V, CV = 70 V) for the IS estriol 3-sulfate. Analytes were separated on a Waters BEH C18 column (1.7 μm, 2.1 × 100 mm) by flowing 5 mM ammonium acetate in water and methanol at 0.4 mL/min. The following gradient was used: 20% organic held for 0.5 min, increased to 45% over 2.5 min, maintained at 45% for 1.5 min, increased to 98% over 1.5 min, and held at 98% for 1 min. The MS peaks were integrated using the QuanLynx software (version 4.1; Waters Corp). For quantification of the hydroxy metabolite, regression fit was based on the analyte/IS peak area ratios calculated from the calibration standards, and the analyte concentration in the incubation was back-calculated using a weighted (1/x 2 ) quadratic least-squares regression (with ±20% error allowed at the low limit of quantification). Hexamutant CYP3A7 (25 mg/ml) in buffer (50 mM potassium phosphate, pH 7.4, 100 mM NaCl, 10% [v/v] glycerol) was saturated with 500 μM DHEA-S. Hexamutant CYP3A7 crystals were grown using sitting drop vapor diffusion equilibration. The hexamutant CYP3A7–DHEA-S complex was equilibrated against 20% (w/v) PEG 3350, 0.2 M lithium acetate dihydrate, pH 7.9 at 25 °C. In contrast to the previous report ( 31 ), sulphate was not required for crystallization and was not observed participating in intermolecular packing with the P21 space group. Crystals formed within 24 h. Crystals were cryoprotected with 70% mother liquor plus 30% glycerol and flash cooled in liquid nitrogen. Diffraction data were collected on beamline 9-2 of the Stanford Synchrotron Radiation Lightsource and processed using XDS ( 51 ). The structure was solved by molecular replacement using Phaser ( 52 ) with a hexamutant CYP3A7 search model (PDB entry: 7MK8 ). Model building and refinement were performed iteratively with Coot ( 53 ) and Phenix.refine ( 54 ), respectively. The ligand DHEA-S was generated using elbow ( 55 ) in Phenix. Figures were generated with PyMOL ( 56 ). Comparisons between structures were generated using the secondary structure matching algorithm ( 57 ) incorporated into Coot ( 53 ). Voids were calculated using VOIDOO ( 58 ) with a probe radius of 1.7 Å.

The authors declare that they have no conflicts of interest with the contents of this article.

All data are included in this article or the supporting information, with the exception of the structure coordinates and structure factors. This latter information has been deposited with the PDB under accession code 8GK3 .

This article contains supporting information .