Three-dimensional structure-function relationship of vitamin D and vitamin D receptor model

Steroids 66 (2001) 177–187

Three-dimensional structure-function relationship of vitamin D and vitamin D receptor model Sachiko Yamada*, Keiko Yamamoto, Hiroyuki Masuno, Mihwa Choi Institute of Biomaterials and Bioengineering, Tokyo Medical and Dental University, 2–3-10 Kanda-Surugadai, Chiyoda-ku, Tokyo 101-0062, Japan

Abstract On the basis of conformational analysis of the vitamin D side chain and studies using conformationally restricted synthetic vitamin D analogs, we have suggested the active space region concept of vitamin D: The vitamin D side-chain region was grouped into four regions (A, G, EA and EG) and the A and EA regions were suggested to be important for vitamin D actions. We extended our theory to known highly potent vitamin D analogs and found a new region F. The analogs which occupy the F region have such modifications as 22-oxa, 22-ene, 16-ene and 18-nor. Altogether, the following relationship between the space region and activity was found: Affinity for vitamin D receptor (VDR), EA ⬎ A⬎ F ⬎ G ⬎ EG; Affinity for vitamin D binding protein (DBP), A ⬎⬎ G,EA,EG; Target gene transactivation, EA ⬎ F ⬎ A ⬎ EG ⭌ G; Cell differentiation, EA ⬎ F ⬎ A ⬎ EG ⭌ G; Bone calcium mobilization, EA ⬎ GA ⬎ F ⭌ EG; Intestinal calcium absorption, EA ⫽ A ⭌ G ⬎⬎ EG. We modeled the 3D structure of VDR-LBD (ligand binding domain) using hRAR␥ as a template, to develop our structure-function theory into a theory involving VDR. 1␣,25(OH)2D3 was docked into the ligand binding pocket of the VDR with the side chain heading the wide cavity at the H-11 site, the A-ring toward the narrow ␤-turn site, and the ␤-face of the CD ring facing H3. Amino acid residues forming hydrogen bonds with the 1␣- and 25-OH groups were specified: S237 and R274 forming a pincer type hydrogen-bond for the 1␣-OH and H397 for the 25-OH. Mutants of several amino acid residues that are hydrogen-bond candidates were prepared and their biologic properties were evaluated. All of our mutation results together with known mutation data support our VDR model docked with the natural ligand. © 2001 Elsevier Science Inc. All rights reserved. Keywords: Nuclear receptor; Vitamin D; Homology modeling; Structure-function relationship

1. Introduction 1␣,25-Dihydroxyvitamin D3, 1␣,25(OH)2D3 (1), is a unique steroid hormone having long and flexible structural features. It should be noted that sunlight irradiation plays a role in producing this flexible seco-steroid. Besides its classic role in regulating calcium metabolism, vitamin D is also involved in such basic functions as regulation of proliferation and differentiation of cells and the immune response [1]. Active vitamin D analogs have been successfully used in the treatment of calcium and bone disorders and the skin disorder psoriasis. However, there is still considerable interest in academia and the pharmaceutical industry to find vitamin D drugs that exhibit specific functions useful in treating immune disorders and malignant tumors. 1␣,25(OH)2D3 exerts these effects through a ligand-activated transcription factor, vitamin D receptor (VDR) [2]. VDR is a member of the nuclear receptor (NR) * Corresponding author. Tel.: ⫹81-3-5280-8036; fax: ⫹81-3-52808039. E-mail address: yamada@i-mde,tmd.ac.jp (S. Yamada).

superfamily [3] to which the receptors for the steroid and thyroid hormones and retinoic acids and numerous orphan receptors belong. To find vitamin D analogs with a specific activity spectrum, it is important to understand (i) the general molecular mechanism of the transactivation mediated by nuclear receptors, (ii) the systematic and theoretical structurefunction relationship of vitamin D, and (iii) the three-dimensional structure of VDR and docking manner of vitamin D ligands. In this review, we will discuss our recent studies related to the second and third subjects, conformation-function relationship of vitamin D side chain [4 – 8] and modeling of VDR-LBD [9].

2. Structure-function relationship of vitamin D side chain Over five hundreds of vitamin D analogs have been synthesized and biologically evaluated [10]. Interestingly, most of highly potent derivatives are modified on the side chain or around the side chain. Thus, we first

0039-128X/01/$ – see front matter © 2001 Elsevier Science Inc. All rights reserved. PII: P I I S 0 0 3 9 - 1 2 8 X ( 0 0 ) 0 0 1 4 5 - 8

178

S. Yamada et al. / Steroids 66 (2001) 177–187

Table 1 Single structural modifications of vitamin D and their effect in biological potency Modification 20-Epimerization Replacement with oxygen Unsaturation

Homologation Methylation Perfluorination Nor-

a

22-Oxa 23-Oxa 22-Ene 16-Ene 23-Yne 24-Homo 24,24-Dihomo 26,27-Me2 20-Me 24-F2 26,27-F6 18-Nor 24-Nor 26,27-Dinor

Effect in potencya

Representative compound

Ref.

20–30 5–10 0.2 1–2 2–5 2–3 5–10 2–5 2–5 7 5–10 10 5–10 0.08 0.01

20-Epi-1␣,25(OH)2D3 (2) 22-Oxa-1␣,25(OH)2D3 (3) 23-Oxa-1␣,25(OH)2D3 (4) 22-Ene-1␣,25(OH)2D3 (5) 16-Ene-1␣,25(OH)2D3 (6) 23-Yne-1␣,25(OH)2D3 (7) 24-Homo-1␣,25(OH)2D3 (8) 24-Dihomo-1␣,25(OH)2D3 (9) 26,27-Me2-1␣,25(OH)2D3 (10) 20-Me-1␣,25(OH)2D3 (11) 24-F2-1␣,25(OH)2D3 (12) 26,27-F6-1␣,25(OH)2D3 (13) 18-Nor-1␣,25(OH)2D3 (14) 24-Nor-1␣,25(OH)2D3 (15) 26,27-Dinor-1␣,25(OH)2D3 (16)

11 12 13 14 15 14 16,17 17 18,19 20 21 22 23 24 24

Cell differentiating activity relative to 1␣,25(OH)2D3 (1), the activity of 1 being defined as 1.

focused our attention on the structure-function relationship of the vitamin D side chain. Structural modifications and their effect on cell differentiating activity are summarized in Table 1 (we use cell differentiating potency as an activity index unless otherwise stated). Among these, what we were most interested in was (i) the epimerization at C(20) that has the largest activity increasing effect (20⬃30-fold) [11]; nonetheless, this modification is expected to change the side chain direction compared with the parent 1␣,25(OH)2D3, and (ii) replacement of C(22) with oxygen which causes dramatic change in the activity pattern: 22-oxa-1␣,25(OH)2D3 (3) has a high cell differentiating activity (10-fold that of the natural hormone) but has little DBP affinity and poor calcemic activity [12]. Substitution of C(23) with oxygen, 23-oxa1␣,25(OH)2D3 (4), just decreases the cell differentiating potency by 1/5 [13]. To answer the question of why these structural changes have such a large effect on the potency, we investigated the structure-activity relationship using conformational analysis as a basic tool. 2.1. Conformational analysis and grouping of the side chain area of 1␣,25(OH)2D3 (1) and 20-epi1␣,25(OH)2D3 (2) The mobility of the side chain of 1␣,25(OH)2D3 (1) and 20-epi-1␣,25(OH)2D3 (2) was analyzed by systematic conformational analysis, and the areas where the 25-oxygen of 1 and 2 can move around are shown on a dot map (Fig. 1a) [5,6]. We then grouped these regions into four: A and G regions occupied by 1␣,25(OH)2D3 (1) and EA and EG regions occupied by 20-epi-1␣,25(OH)2D3 (2). The terms A and G refer to anti and gauche and E to epi: 1 occupies the A and G regions when its dihedral angle at C(17–20-22–23) is anti and gauche(⫹), respectively, while 2 occupies the EA and EG regions when its dihedral angle at C(17–20-22– 23) is anti and gauche(-), respectively.

2.2. Relationship between the side-chain region and activities: biologic activities of side-chain restricted analogs To determine the relationship between these side chain regions and activity, we designed and synthesized analogs whose side chain mobility is confined to one of the four regions, A, G, EA and EG, four diastereomers (17– 20) at C(20) and C(22) of 22-Me-1␣,25(OH)2D3 [4 – 6]. The biologic activities of these analogs (17–20) were then evaluated compared with the natural hormone 1 (Table 2) [8]. 2.2.1. Affinity for VDR The affinity of the four diastereomers of 22-Me1␣,25(OH)2D3 (17–20) for the porcine intestinal and the bovine thymus VDRs was evaluated. The pairs of methylated analogs, 17 and 18 derived from 1␣,25(OH)2D3 (1), and 19 and 20 from 20-epi-1␣,25(OH)2D3 (2), showed distinct affinities. In the porcine VDR assay, the potency of 22S-Me-1␣,25(OH)2D3 (18) was 1/3 of that of 1, whereas the 22R-isomer 17 was 1/60 of the potency of 1. A very striking difference was found between the pair of epimers (19 and 20) with 20-epi-configuration. In the porcine VDR assay, the 22R-epimer 19 had 20 times higher affinity than 1, whereas the 22S-isomer 20 was 1/100 of the potency of 1. The results with thymus VDR were similar. Thus, the affinity ratio for the pair of conformationally restricted analogs (18/17) of 1 was about 20, whereas that for the analogs (19/20) of 2 was about 2000. These results indicate that the A and EA regions are important for VDR binding. Thus, in terms of the side-chain region, order of the VDR affinity was EA ⬎ A ⬎ G ⬎ EG. 2.2.2. Affinity for vitamin D binding protein (DBP) The affinity for DBP was tested using the vitamin Ddeficient rat serum DBP. Only 22S-Me-1␣,25(OH)2D3 (18)

S. Yamada et al. / Steroids 66 (2001) 177–187

179

Fig. 1. (a) Mobile areas of 25-OH group of 1 and 2 and groups of regions: A and G regions of 1 and EA and EG regions of 2 (stereoview). (b-j) Side chain areas of vitamin D analogs (crosses) compared with the four regions, A, G, EA and EG, of 1 and 2 (dots) (stereoviews): (b) 22-Oxa-1␣,25(OH)2D3 (3), (c) 22-ene-1␣,25(OH)2D3 (5), (d) 22,24-diene-24,26,27-trihomo-1␣,25(OH)2D3 (21), (e) 16-ene-1␣,25(OH)2D3 (6), (f) 18-nor-1␣,25(OH)2D3 (14), (g) 16-en23-yne-1␣,25(OH)2D3 (22), (h) 20-epi-16-en-23-yne-1␣,25(OH)2D3(23), (i) (17Z)- (25, left crosses) and (17E)-en-22-yne-24,26,27-trihomo-1␣,25-(OH)2D3 (24, right crosses), (j) MC903 (26).

180

S. Yamada et al. / Steroids 66 (2001) 177–187

Table 2 Side chain region and various activitiesa Compd.

1 17 18 19 20 2

Side-chain region

A&G G A EA EG EA & EG

Binding affinity VDR (porcine)

VDR (bovine)

DBP (rat)

1 0.02 0.33 20 0.01 –

1 0.02 0.33 11 ⬍0.01 5

1 ⬍0.01 0.67 ⬍0.01 ⬍0.01 –

Gene trans. OPNf

Cell diff. HL-60

In vivo activity BCM

ICA

1 0.01 0.22 100 0.03 40

1 0.01 1 120 0.08 –

1 0.32 0.26 2.20 0.05 –

1 0.25 0.50 0.51 ⬍0.01 –

a Activity relative to 1␣,25(OH)2D3 1, activity of 1 being defined as 1 (the larger the value, the higher the activity). OPN, osteopontin; BCM, bone calcium mobilization; ICA, intestinal calcium absorption.

showed significant affinity (2/3 as active as 1) indicating that the region responsible for DBP binding is only A. 2.2.3. Transcription activity In a transient transfection assay using mouse osteopontin (OPN) responsive element, isomer 19 in the EA region showed the highest potency (100 times as active as 1) and 18 in the A region was next, in accord with the VDR affinity. The pairs of the 22-epimers (18/17 and 19/20) again exhibited distinct activities, but the potency difference was much more striking between the isomers (19/20) with 20epi configuration: the activity ratio of 19/20 is 3000 and that of 18/17 is 22. It should be noted that, in 20-epi compounds (19 and 20), the transcriptional activity is amplified (3–9fold) compared with VDR affinity. Thus, the order of transcriptional activity in terms of the spatial region was EA ⬎ A ⬎ EG ⭌ G. 2.2.4. In vitro activity. Differentiation of HL-60 cells Cell differentiating activity was evaluated using human promyelocytic leukemia (HL-60) cells. Isomer 19 in the EA region has the highest potency (100 times as active as 1). The potency difference between the 22-epimer pairs was again obvious: activity ratio of 18/17 was 100 and that of 19/20 was 1500. Order of cell differentiation activity was EA ⬎ A ⬎ EG ⭌ G, the same as the order of transcriptional potency. 2.2.5. In vivo activity. Bone calcium mobilization (BCM) and intestinal calcium transport (ICA) In vivo calcemic activities BCM and ICA were evaluated. Notably, 19 in EA showed the highest potency for in vivo BCM, nonetheless it has a low affinity for the transport protein (DBP) of vitamin D. However, the same was not true in ICA. Thus DBP affinity may probably be necessary for in vivo ICA activity. DBP is believed to be important for ensuring that vitamin D has a long survival time in the blood. Thus, a long period of survival in the blood is likely necessary for ICA activity but not for BCM. It is also interesting that the 22-epimers 17 and 18 of 20R-analogs show similar in vivo calcemic activity, although the two

isomers showed distinct potency in the gene transcription and the cell differentiation assays. In terms of the side chain region, the orders of potency are BCM: EA ⬎ G ⭌ A ⬎ EG and ICA: EA ⫽ A ⭌ G ⬎⬎ EG. 2.3. Side-chain region and activity relationship in other highly potent vitamin D analogs We analyzed the conformation of over 50 highly potent 1␣,25(OH)2D3 derivatives [7] by a method similar to that described above [6]. The results support our active space group concept with only a few exceptions. We also defined a new region F that is placed in front of the EA to A regions. 2.3.1. Compounds with a side chain at the F region We found a group of compounds whose 25-oxygen occupies an area in front of the EA to A regions. We termed this area F. The compounds which occupy this area have 22-oxa-, 16-ene-, 22-ene- or 18-nor modifications, such as 22-oxa-1␣,25(OH)2D3 (3) (Fig. 1b), 22-ene-1␣,25(OH)2D3 (5) (Fig. 1c), 22,24-diene-24,26,27-trihomo-1␣,25(OH)2D3 (21) (Fig. 1d) [25], 16-ene-1␣,25(OH)2D3 (6) (Fig. 1e), and 18-nor-1␣,25(OH)2D3 (14) (Fig. 1f). 22-Oxa-1␣,25(OH)2D3 (3) was the first compound in which the cell differentiating and calcemic activities of vitamin D were separated [12]. It has not been explained in structural terms why replacing C(22)H2 with oxygen dramatically changes the potency spectrum. Conformational analysis provided one explanation. Replacing C(22)H2 with oxygen changes the minimum energy conformation of the side chain and, in turn, the area occupied by the 25-oxygen. The C(22)H2 and 22-oxa compounds differ in their C(16 – 17-20 –22) torsion angle in the minimum-energy conformation. The C(22)H2 compound adopts a C(16 –17-20 –22) gauche(⫹) conformation, whereas the 22-oxa compound takes a gauche(-) conformation. In the gauche(-) conformation, C(22)H2 suffers severe steric repulsion from the C(18)H3 group. However, in the 22-oxa analog, this repulsion is reduced because the oxygen bears no hydrogen, so it can adopt a C(16 –17-20 –22) gauche(-) conformation. When vitamin D adopts a C(16 –17-20 –22) gauche(⫹), the

S. Yamada et al. / Steroids 66 (2001) 177–187

181

Fig. 2. Sequence alignment of hVDR-LBD with hRAR␥-LBD. Shadows indicate identical residues; Circles represent the residues facing LBP. Bars indicate regions of secondary structures.

25-oxygen tends to reside in the A and G regions, whereas when it adopts a gauche(-) the 25-oxygen tends to locate in the F region. For the same reason, ⌬22 compounds, such as 22-ene-1␣,25(OH)2D3 (5) (Fig. 1c) and 22,24-diene24,26,27-trihomo-1␣,25(OH)2D3 (21) (Fig. 1d), occupy the F region. The result indicates that global minimum energy conformation is important in the expression of biologic potency. 16-Ene-1␣,25(OH)2D3 (6) and its derivatives also occupy the F region. The high potency of these vitamin D analogs developed by the Roche group [26,27] has long been known without any explanation. We found that the introduction of a double bond to C(16) also directs the side chain to the F region (Fig. 1e). In the minimum-energy conformation, C(16 –17-20 –22) dihedral angle of 16-ene1␣,25(OH)2D3 (6) is ⫺54°, similar to the same dihedral angle of 3. Introduction of a triple bond to C(23) of 6 yields analog 22 with increased potency (10-fold) [26,27]. However epimerization at C(20) of 22 yielding 20-epi-16-en-23yne-1␣,25(OH)2D3 (23) [27] reduces the potency by 1/20. This effect can also be explained by comparing the dot maps of 22 and 23 (Fig. 1g and 1h): While the dots of 22 are found over the active front regions of A, F and EA, those of 23 are distributed over the inactive rear regions of EG to G. 18-Nor-1␣,25(OH)2D3 (14) occupies mainly the F region. Removal of the angular C(18)H3 group eliminates congestion between the methyl group and C(22)H2. The side chain therefore can rotate nearly freely around the 17,20 bond, and the 25-oxygen can occupy over a wide range. However, when 20% stable conformations are selected, the 25-oxygen is highly concentrated in the F region (Fig. 1f). Thus, we explain the high potency of 14 [23] because it occupies the F region. 20-Methylation of 1␣,25(OH)2D3 elevates the potency by 7.1 times [20]. Conformational analysis of 20-Me-1␣,25(OH)2D3 (11) explains this effect. This modification makes the three staggered conformations at C(16 –17-20 –22) of similar energy, the steric energy increasing in the order anti ⬍ gauche(-) ⬍ gauche(⫹). Thus, the dots of 11 are distributed over the EA, EG and F regions. Activity orders including the F region are summarized as follows: VDR affinity, EA ⬎ A ⬎ F ⬎ G ⬎ EG; DBP affinity, only A; Target gene transactivation, EA ⬎ F ⬎ A ⬎ EG ⭌ G; Cell differentiation, EA ⬎ F ⬎ A ⬎ EG ⭌ G; In vivo BCM, EA ⬎ G ⭌ A ⬎ F ⭌ EG. Except for in vivo calcemic activity, vitamin D compounds are more

potent when their side chain is located in the front regions than in the rear regions, and in the left regions than in the right regions (when the structures are drawn as shown in Fig. 1). Stereoisomers 22 and 23 with a double bond at C(16) are a typical example of the former situation (Fig. 1g and 1h) and geometrical isomers 24 and 25 at the 17(20)-double bond are the best examples of the latter (Fig. 1i). 17Z-Isomer (25) is 280-fold more potent than the 17E- isomer (24): Potency relative to 1 is 710 and 2.5, respectively [28]. Compounds with discriminated vitamin D actions, the compounds with high cell differentiating and low calcemic activity, are most frequently found in compounds in the F region. In accord with this, many analogs in the F region, such as 26 (MC903, Fig. 1j) [29], 3 (OCT, Fig. 1b), 22 (Ro.23–7553, Fig. 1g) and 21 (EB1089, Fig. 1d), are used or being developed as non-calcemic vitamin D agents. 2.3.2. Modifications not related to side chain conformation Elongation of the side chain increases the potency 5–10 fold in both 20-normal [16] and 20-epivitamin D analogs [11,28]. Also methylation of the terminal 26- and 27-positions of the side chain elevates potency 25-fold [19]. The most likely explanation for the effect of these modifications is increased hydrophobic interaction between the ligands and VDR. Truncation of the side chain, on the other hand, causes significant reduction of potency, as the cases of 24-nor1␣,25(OH)2D3 (15) [24] and 26,27-dinor-1␣,25(OH)2D3 (16) [24] indicate. These examples indicate the importance of the hydrophobic interaction to anchor a ligand on the receptor. Perfluorination at C(24) (12) [21] or C(26 and 27) (13) [22] elevates potency both in the cell differentiating and in the calcemic actions. The effect of perfluorination may also be ascribed to increased hydrophobic interaction between the ligand and VDR.

3. Modeling of vitamin D receptor ligand-binding domain (LBD) and structure-function studies using the model A three-dimensional structure of VDR-LBD is essential to understand the structure-function relationship of vitamin D. Recently the X-ray crystal structures of a number of

182

S. Yamada et al. / Steroids 66 (2001) 177–187

Fig. 3. (a) Ribbon-tube presentation of the VDR-LBD model. (b) Amino acid residues forming the ligand binding cavity and their interaction with 1␣,25(OH)2D3. (c) The residues interacting with 1␣,25(OH)2D3 (stereoviews).

NR-LBDs have been solved [30]. However, the crystal structure of VDR-LBD has not been solved. Two computer models of VDR-LBD have been reported [31,32] but none of them was substantiated experimentally by mutation study, etc. We modeled VDR-LBD harboring the natural

hormone computationally using homology modeling technique [9]. Furthermore we substantiated the model by mutating several amino acid residues assumed to interact with the ligand and evaluating the effect of the mutation on the biologic potency.

S. Yamada et al. / Steroids 66 (2001) 177–187

183

Fig. 4. (a) Overlay of ligands in various NRs: estradiol (E2) in ER␣, progesterone in PR, and all-trans-retinoic acid in RAR␥. (b) Overlay of ER ligands in ER␣: E2, diethylstilbestrol, 4-hydroxytamoxifen and raloxifene. (c) Overlay of the three ligands shown in (a) plus 1␣,25(OH)2D3 (1) docked in our VDR-LBD model.

3.1. Homology modeling Sequence alignment is important in homology modeling. We aligned the sequence of hVDR-LBD (residues 124 – 427) as shown in Fig. 2. We used the X-ray crystal structure of hRAR␥ [33] as a template for the reasons: (i) RAR is a member of the same NR-subfamily (group 1) [34] as VDR and has the highest homology with VDR (identity, 25%; similarity, 45%, excluding the loop between H1 and H3, loop 1–3). (ii) The molecular shape of the natural VDR ligand 1␣,25(OH)2D3 somewhat resembles that of the RAR ligand retinoic acid. A big difference between VDR-LBD and RAR-LBD is the insertion of the long loop 1–3 in the VDR. We modeled VDR by eliminating the loop 1–3 (143–223). Most of the C␣ framework of VDR-LBD was treated as a structurally conserved region and was constructed mimicking as much as possible the framework of the template. The side chains were added, and the conformation of the side chains was checked one by one and manually modified. After optimization, the ligand was docked manually and

optimized again. The resulting model shown in Fig. 3a satisfied the evaluation by the PROCHECK program. The ligand binding pocket (LBP) is located between H11 and the ␤-turn surrounded by H3–7 and H12. The cavity is wide at site 2 and narrow at site 1: Generally, there are two ligand-anchoring sites at the two extremities of NR-LBP, one facing the ␤-turn called site 1 and the other facing H11 called site 2 (notation by Moras, [35]). The cavity is mostly lined with hydrophobic residues (Figs. 2 and 3b), but hydrophilic residues are exposed at the two extremities (S237, K240, R274, S275, S278 and C288 at site 1; H397 and Q400 at site 2). These hydrophilic residues are potential interaction sites for the 1␣-, 3␤- and 25-hydroxyl groups of 1␣,25(OH)2D3. 3.2. Ligand docking At first, we investigated the docking mode of ligands in known 3D structures of NR/ligand complexes. Thus, NR-ligand complexes were overlaid by allowing their C␣ to coincide at the signature region, and then the protein

184

S. Yamada et al. / Steroids 66 (2001) 177–187

Fig. 5. Binding capability and transcriptional activity of the WT and the mutant VDRs. (a) Specific binding to 1␣,25(OH)2[3H]D3. The WT and the mutant hVDRs were synthesized in vitro in a rabbit reticulocyte lysate. The lysate was incubated with increasing concentrations of 1␣,25(OH)2[3H]D3 for 16 h at 4°C. Bound and unbound ligands were separated by dextran-coated charcoal. (b) Transcriptional activity. Cos-7 cells were cotransfected with WT or mutant hVDR expression vectors, SPPx3-TK-Luc as a reporter plasmid and pRL-CMV vector as an internal control. Before harvesting, the cells were treated for 16 h with 10⫺8 M 1␣,25(OH)2D3. Transactivation was determined by luciferase activity and normalized to the internal control.

parts were deleted (Fig. 4a) [36 –39]. Interestingly, all the ligands are harbored in the same region of the NRs. The ligands are aligned at site 2 and are variable depending on the length of the ligand at site 1. However, when various ligands in a particular NR are compared, the positions of the ligands are variable at site 2 and are stringent at site 1 as various ER ligands in hER␣ show (Fig. 4b) [37,38, 40]. In docking vitamin D into the LBP, four orientations of 1␣,25(OH)2D3 are possible. With respect to the long molecular axis, we chose the orientation in which the side chain part is accommodated in the wide cavity at site 2 and the A-ring in the narrow cavity at site 1. This satisfies the structure-function relationship of vitamin D that variety in the side chain structures is acceptable in VDR, while only a limited structural modification is allowed in the A-ring. Introduction of substituents to the A-ring causes a drastic reduction in the potency, except for the 2-position [41,42]. With respect to the short axis, we chose the ␤-face-up orientation, the same orientation as that of estradiol [37,38,40] and progesterone [38,39] in their receptor complexes. Having determined the direction of the ligand, we overlaid 1␣,25(OH)2D3 with other NR-ligands (Fig. 4c), docked it manually in the LBP and then searched for hydrogen bond partners of the three hydroxyl groups computationally. The 1␣-hydroxyl group forms pincer-type hydrogen bonds with S237 (hydrogen acceptor) and R274 (hydrogen donor) (Fig. 3c). Pincer-type hydrogen bonds are commonly observed at site 1 in the crystal structures of the NR/natural ligand complexes and appear to play an important role in rigidly anchoring ligands in their LBP. The 25-hydroxyl group forms a hydrogen bond with the imidazole ring of H397 (H11) which acts as the hydrogen donor. We were not

able to specify an amino acid residue interacting with the 3␤-hydroxyl group of the ligand. The 3␤-hydroxyl group may not form a hydrogen bond with the amino acid inside VDR-LBP, because the removal of either the 1␣- or 25-hydroxyl group causes a significant reduction (1/500 – 1/1000) in the affinity for VDR, whereas the removal of the 3␤-hydroxyl group has a much smaller effect (1/17) [10]. Ray et al. reported in this meeting (Abstract of Papers, First International Conference on Chemistry and Biology of Vitamin D Analogs, Providence, September, 1999) that affinity labeling with 1␣,25(OH)2D3 3␤-bromoacetate occurred exclusively at C288. The data support the ␤-face-up orientation of the ligand in our model. As described below, mutant C288A showed significantly reduced activities indicating that C288 has some role in binding the ligand. 3.3. Mutation analysis Our 3D model of the VDR-LBD/1␣,25(OH)2D3 complex was substantiated by mutation analysis. Five one point mutants, S237A, S275A, S278A, C288A, and H397A, were prepared and their biologic properties were evaluated. The mutant of R274 (R274L) has been known in a human type II rickets and was confirmed to lack activity in both ligand binding and transactivation [43]. As shown in Fig. 5a, the ability to bind to the natural ligand was completely abolished in H397A, significantly reduced in S237A and C288A and little affected in S275A and S278A. The transcription activity of the five mutant VDRs was in parallel with their binding affinity for 1␣,25(OH)2D3 (Fig. 5b). Thus it is certain that H397 and R274 play a crucial role and that S237 and C288 have significant role in ligand binding. Thus the mutation data firmly support our model.

S. Yamada et al. / Steroids 66 (2001) 177–187

3.4. Surface structure Transactivation function 2 (AF-2) surface was successfully created at a position similar to that of RAR, and the residues forming charge clump (E420 at helix 12 and K246 at helix 3) were perfectly positioned at the opposite ends of the hydrophobic cleft of the AF-2 surface supporting the quality of our model [9]. 3.5. Docking of other ligands and explanation of their activity We placed the dot map of the vitamin D side chain (shown in Fig. 1b) in the LBP of our VDR model and investigated the relationship between the active side-chain regions and the 3D structure of LBP [9]. The active regions of A, EA and F can be harbored in the wide cavity at site 2, and the 25-OH in these regions can form a hydrogen bond with H397 if it rotates its ␹2, while the 25-OH group in the G and EG regions is placed in a region where it cannot form a hydrogen bond with H397. It has been reported in ER/ ligand complexes that H524 at H11 forms hydrogen bond with different ligands by rotating its ␹2. Thus, we conclude that the vitamin D compounds whose side chain is in the A, EA or the F region are biologically potent because their 25-OH group can form a hydrogen bond with H397. The high potency of 20-epivitamin D analogs can be explained as follows: The side chain of the 20-epivitamin D analogs is directed toward H12 and the terminal methyl groups can interact with the hydrophobic residues inside H12: This hydrophobic interaction stabilizes the transcriptionally active form of VDR where H12 covers LBP forming the AF-2 surface. The VDR affinity of 20-epi analogs with longer side chain, such as KH1060 (27), is

185

lower than that of the natural hormone (1), but its in vitro activity is much higher than that of 1 [11]. The long side chain of 27 may have a much stronger hydrophobic interaction with H12 stabilizing the active form of VDR.

4. Conclusions While preparing this review article, the X-ray crystal structure of a mutant VDR-LBD, in which the thermally labile loop 1–3 was engineered to be crystallized (50 amino acid residues were eliminated from the loop 1–3), was reported [44]. Our VDR model is quite similar to the crystal structure: residues which form hydrogen bonds with 25-OH (H397) and 1␣-OH (S237 and R274, pincer-type) and the orientation of vitamin D in the LBP were the same with ours. Obvious differences are (i) that the 3␤-OH group forms a hydrogen bond with S278 in the crystal structure, though we could not specify the hydrogen-bond partner for 3␤-OH. In our mutation analysis, however, S278A had little effect on the ligand binding and on transcriptional activity. (ii) The A-ring adopts a ␤-form in the crystal structure, though we adopted an ␣-conformation in our model. In the mutant VDR, the loop 1–3 is placed between the ␤-turn and H3, as is the loop 1–3 of NRs (ER and PR) belonging to subgroup 3 [37– 40]. Because, the loop 1–3 is placed outside the ␤-turn in the NRs of subgroup 1 to which VDR belongs [9,36], the structure of the mutant VDR-LBP may probably be different from that of wild-type VDR-LBP at site 1. Wild-type VDR and its one point mutants exhibit different responses to various vitamin D analogs. Studies using these mutants would provide valuable information about the conformation of ligands inside VDR/LBP. Structure-based design and synthesis of non-steroid vitamin D are now

186

S. Yamada et al. / Steroids 66 (2001) 177–187

possible in the vitamin D field. These studies are now progressing in our laboratory.

[17]

References [18] [1] Feldman D, Glorieux FH, Pike JW, editors. Vitamin D. San Diego: Academic Press, 1997. [2] DeLuca HF. Mechanism of action of 1␣,25-dihydroxyvitamin D3. J Bone Miner Metab 1990;8:1–9. [3] Mangelsdorf DJ, Thummel C, Beato M, Herrlich P, Schutz G, Umesono K, Blumberg B, Kastner P, Mark M, Chambon P, Evans RM. The nuclear receptor superfamily: the second decade. Cell 1995;83: 835–9. [4] Yamamoto K, Takahashi J, Hamano K, Yamada S, Yamaguchi K, DeLuca HF. Stereoselective syntheses of (22R)- and (22S)-22-methyl-1␣,25-dihydroxyvitamin D3: Active vitamin D3 analogs with restricted side chain conformation. J Org Chem 1993;58:2530 –7. [5] Yamamoto K, Ohta M, DeLuca HF, Yamada S. On the side chain conformation of 1␣,25-dihydroxyvitamin D3 responsible for binding to the receptor. Bioorg Med Chem Lett 1995;5:979 – 84. [6] Yamamoto K, Sun W-Y, Ohta M, Hamada K, DeLuca HF, Yamada S. Conformationally restricted analogs of 1␣,25-dihydroxyvitamin D3 and its 20-epimer: Compounds for study of the three-dimensional structure of vitamin D responsible for binding to the receptor. J Med Chem 1996;39:2727–37. [7] Yamada S, Yamamoto K, Masuno H, Ohta M. Conformation-function relationship of vitamin D: Conformational analysis predicts potential side chain structure. J Med Chem 1998;41:1467–75. [8] Yamamoto K, Ooizumi H, Umesono K, Verstuyf A, Bouillon R, DeLuca HF, Shinki T, Suda T, Yamada S. Three-dimensional structure-function relationship of vitamin D: side chain location and various activities. Bioorg Med Chem Lett 1999;9:1041– 6. [9] Yamamoto K, Masuno H, Choi M, Nakashima K, Taga T, Ooizumi H, Umesono K, Sicinska W, VanHooke J, DeLuca HF, Yamada S. Three-dimensional modeling of and ligand docking to vitamin D receptor ligand-binding domain. Proc Natl Acad Sci USA 2000;97: 1467–72. [10] Bouillon R, Okamura WH, Norman AW. Structure-function relationships in the vitamin D endocrine system. Endocr Rev 1995;16:200 – 57. [11] Binderup L, Latini S, Binderup E, Bretting C, Calverley M, Hansen K. 20-Epi-vitamin D3 analogues: A novel class of potent regulators of cell growth and immune responses. Biochem Pharmacol 1991;42: 1569 –75. [12] Abe J, Takita Y, Nakano T, Miyaura C, Suda T, Nishii Y. A synthetic analogue of vitamin D3, 22-oxa-1␣,25-dihydroxyvitamin D3, is a potent modulator of in vivo immunoregulating activity without inducing hypercalcemia in mice. Endocrinology 1989;124:2645–7. [13] Kubodera N, Miyamoto K, Akiyama M, Matsumoto M, Mori T. Synthetic studies of vitamin D analogues. Synthesis and differentiation-inducing activity of 1␣,25-dihydroxy-23-oxa, thia-, and azavitamin D3. Chem Pharm Bull 1991;39:3221– 4. [14] Chen TC, Persons K, Uskokovic MR, Horst RL, Holick MF. An evaluation of 1,25-dihydroxyvitamin D3 analogues on the proliferation and differentiation of cultured human keratinocytes, calcium metabolism and the differentiation of human HL-60 cells. J Nutr Biochem 1993;4:49 –57. [15] Zhou JY, Norman AW, Akashi M, Chen DL, Uskokovic MR, Aurrecoechea JM, Dauben WG, Okamura WH, Koeffler HP. Development of a novel 1␣,25(OH)2-vitamin D3 analog with potent ability to induce HL-60 cell differentiation without modulating calcium metabolism. Blood 1991;78:75– 82. [16] Ostrem VK, Tanaka Y, Prahl J, DeLuca HF, Ikekawa N. 24- and 26-homo-1␣,25-dihydroxyvitamin D3: preferential activity in induc-

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

[30] [31]

ing differentiation of human leukemia cells HL-60 in vitro. Proc Natl Acad Sci USA 1987;84:2610 – 4. Perlman K, Kutner A, Prahl J, Smith C, Inaba M, Schnoes HK, DeLuca HF. 24-Homologated 1␣,25-dihydroxyvitamin D3 compounds: separation of calcium and cell differentiation activities. Biochemistry 1990;29:190 – 6. Eguchi T, Ikekawa N, Sumitani K, Kumegawa M, Higuchi S, Otomo S. Effect on carbon lengthening at the side chain terminal of 1␣,25dihydroxyvitamin D3 for calcium regulating activity. Chem Pharm Bull 1990;38:1246 –9. Honda A, Mori Y, Otomo S, Ishizuka S, Ikekawa N. Effects of novel 26,27-dialkyl analogs of 1␣,25-dihydroxyvitamin D3 on differentiation-inducing activity of human promyelocytic leukemia (HL-60) cells in serum-supplemented or serum-free culture. Steroid 1991;56: 142–7. Danielsson C, Nayeri S, Wiesinger H, Thieroff-Ekerdt R, Carlberg C. Potent gene regulatory and antiproliferative activities of 20-methyl analogues of 1␣,25 dihydroxyvitamin D3. J Cellular Biochem 1996; 63:199 –206. Shiina Y, Abe E, Miyaura C, Tanaka H, Yamada S, Ohmori M, Nakayama K, Takayama H, Matsunaga I, Nishii Y, DeLuca HF, Suda T. Biological activity of 24,24-difluoro-1␣,25-dihydroxyvitamin D3 and 1␣,25-dihydroxyvitamin D3 26,23-lactone in inducing differentiation of human myeloid leukemia cells. Arch Biochem Biophys 1983;220:90 – 4. Inaba M, Okuno S, Nishizawa Y, Yukioka K, Otani S, Matsui-Yuasa I, Morisawa S, DeLuca HF, Morii H. Biological activity of fluorinated vitamin D analogs at C-26 and C-27 on human promyelocytic leukemia cells, HL-60. Arch Biochem Biophys 1987;258:421–5. Sicinski RR, Perlman KL, Prahl J, Smith C, DeLuca HF. Synthesis and biological activity of 1␣,25-dihydroxy-18-norvitamin D3 and 1␣,25-dihydroxy-18,19-dinorvitamin D3. J Med Chem 1996;39: 4497–506. Ostrem VK, Lau WF, Lee SH, Perlman K, Prahl J, Schnoes HK, DeLuca HF. Induction of monocytic differentiation of HL-60 cells by 1␣,25-dihydroxyvitamin D analogs. J Biol Chem 1987;262:14164 – 71. Binderup E, Calverley MJ, Binderup L. Synthesis and biological activity of 1␣-hydroxylated vitamin D analogues with poly-unsaturated side chains. In: Norman AW, Bouillon R, Thomasset M, editors. Vitamin D: Gene Regulation, Structure-Function Analysis and Clinical Application. Berlin: Walter de Gruyter, 1991. pp. 192–3. Uskokovic MR, Baggiolini E, Shiuey SJ, Jacobelli J, Hennessy B, Kiegiel J, Daniewski AR, Pizzolato G, Coustney LF, Horst RL. The 16-ene-analogs of 1␣,25-dihydroxycholecalciferol synthesis and biological activity. In: Norman AW, Bouillon R, Thomasset M, editors. Vitamin D: Gene Regulation, Structure-Function Analysis and Clinical Application. Berlin: Walter de Gruyter, 1991. pp. 139 – 45. Uskokovic MR, Studzinski GP, Gardner JP, Reddy GS, Campbell MJ, Koeffler HP. The 16-ene vitamin D analogs. Curr Pharm Des 1997;3:99 –123. Bretting C, Hansen CM, Andersen NR. Chemistry and biology of 22,23-yne analogues of calcitiol. In: Norman AW, Bouillon R, Thomasset M, editors. Vitamin D, a Pluripotent Steroid Hormone: Structural Studies, Molecular Endocrinology and Clinical Applications. Berlin, New York: Walter de Gruyter, 1994. pp. 73– 4. Binderup L, Bramm E. Effects of a novel vitamin D analogue MC903 on cell proliferation and differentiation in vitro and on calcium metabolism in vivo. Biochem Pharmacol 1988;37:889 –95. Weatherman RV, Fletterick RJ, Scanlan TS. Nuclear-receptor ligands and ligand-binding domains. Annu Rev Biochem 1999;68:559 – 81. Wurtz J-M, Guillot B, Moras D. 3D model of the ligand binding domain of the vitamin D nuclear receptor based on the crystal structure of holo RAR␥. In: Norman AW, Bouillon R, Thomasset M, editors. Vitamin D: chemistry, biology and clinical applications of the

S. Yamada et al. / Steroids 66 (2001) 177–187

[32]

[33]

[34] [35]

[36]

[37]

[38]

steroid hormone. Riverside, CA: University of California-Riverside Printing and Reprographics, 1997. pp. 165–72. Norman AW, Adams D, Collins ED, Okamura WH, Fletterick RJ. Three-dimensional model of the ligand binding domain of the nuclear receptor for 1␣,25-dihydroxyvitamin D3. J Cellular Biochem 1999; 74:323–33. Renaud JP, Rochel N, Ruff M, Vivat V, Chambon P, Gronemeyer H, Moras D. Crystal structure of the RAR-␥ ligand-binding domain bound to all-trans retinoic acid. Nature 1995;378:681–9. Laudet V, Hanni C, Coll J, Catzeflis F, Stehelin D. Evolution of the nuclear receptor gene superfamily. EMBO J 1992;11:1003–13. Fagart J, Wurtz JM, Souque, Hellal-Levy C, Moras D, Rafestin-Oblin ME. Antagonism in the human mineralocorticoid receptor. EMBO J 1998;17:3317–25. Renaud JP, Rochel N, Ruff M, Vivat V, Chambon P, Gronemeyer H, Moras D. Crystal structure of the RAR-␥ ligand-binding domain bound to all-trans retinoic acid. Nature 1995;378:681–9. Brzozowski AM, Pike AC, Dauter Z, Hubbard R, Bonn T, Engstrom O, Ohman L, Greene GL, Gustafsson JA, Carlquist M. Molecular basis of agonism and antagonism in the oestrogen receptor. Nature 1997;389:753– 8. Tanenbaum DM, Wang Y, Williams SP, Sigler PB. Crystallographic comparison of the estrogen and progesterone receptor’s ligand binding domains. Proc Nat Acad Sci USA 1998;95:5998 – 6003.

187

[39] Williams SP, Sigler PB. Atomic structure of progesterone complexed with its receptor. Nature 1998;393:392– 6. [40] Shiau AK, Barstad D, Loria PM, Cheng L, Kushner PJ, Agard DA, Greene GL. The structural basis of estrogen receptor/coactivator recognition and the antagonism of this interaction by tamoxifen. Cell 1998;95:927–37. [41] Konno K, Maki S, Fujishima T, Liu ZP, Miura D, Chokki M, Takayama H. A novel and practical route to A-ring enyne synthon for 1␣,25-dihydroxyvitamin D3 analogs: synthesis of A-ring diastereomers of 1␣,25-dihydroxyvitamin D3 and 3-methyl-1␣,25-dihydroxyvitamin D3. Bioorg Med Chem Lett 1998;8:151– 6. [42] Ono Y, Watanabe H, Shiraishi A, Takeda S, Higuchi Y, Sato K, Tsugawa N, Okano T, Kobayashi T, Kubodera N. Synthetic studies of vitamin D analogs. XXIV. Synthesis of active vitamin D3 analogs substituted at the 2␤-position and their preventive effects on bone mineral loss in ovariectomized rats. Chem Pharm Bull 1997;45: 1626 –30. [43] Kristjansson K, Rut AR, Hewison M, O’Riordan JLH, Hughes MR. Two mutations in the hormone binding domain of the vitamin D receptor cause tissue resistance to 1␣,25-dihydroxyvitamin D3. J Clin Invest 1993;92:12– 6. [44] Rochel N, Wurtz JM, Mitschler A, Klaholz B, Moras D. The crystal structure of the nuclear receptor for vitamin D bound to its natural ligand. Mol Cell 2000;5:173–9.