Promiscuity and diversity in 3-ketosteroid reductases

Abstract

Many steroid hormones contain a Δ⁴-3-ketosteroid functionality that undergoes sequential reduction by 5α- or 5β- steroid reductases to produce 5α- or 5β-dihydrosteroids; and a subsequent 3-keto-reduction to produce a series of isomeric tetrahydrosteroids. Apart from steroid 5α-reductase all the remaining enzymes involved in the two step reduction process in humans belong to the aldo-keto reductase (AKR) superfamily. The enzymes involved in 3-ketosteroid reduction are AKR1C1–AKR1C4. These enzymes are promiscuous and also catalyze 20-keto- and 17-keto-steroid reduction. Interest in these reactions exist since they regulate steroid hormone metabolism in the liver, and in steroid target tissues, they may regulate steroid hormone receptor occupancy. In addition many of the dihydrosteroids are not biologically inert. The same enzymes are also involved in the metabolism of synthetic steroids e.g., hormone replacement therapeutics, contraceptive agents and inhaled glucocorticoids, and may regulate drug efficacy at their cognate receptors. This article reviews these reactions and the structural basis for substrate diversity in AKR1C1–AKR1C4, ketosteroid reductases.

This article is part of a Special Issue entitled ‘Steroid/Sterol signaling’.

1. Introduction

The majority of steroid hormones and their synthetic derivatives contain a Δ⁴-3-ketosteroid functionality in their A-ring. Since the early work of Tomkins [1,2], it has been recognized that when this group is present it undergoes 5α- or 5β-reduction catalyzed by either 5α-reductase enzymes (SRD5A1, SRD5A2 and SRD5A3) [3,4] or by steroid 5β-reductase [5,6], respectively to produce the corresponding 5α/5β-dihydrosteroids, Fig. 1. These 5α/5β-dihydrosteroids are then reduced by NADPH-dependent 3α/3β-hydroxysteroid dehydrogenases to produce one of four stereoisomeric tetrahydrosteroids. Once formed the tetrahydrosteroids undergo conjugation reactions with either SULTs (sulfotransferases) or UGTs (uridine glucuronosyl transferases). The functionalization of the A-ring to yield the four isomeric tetrahydrosteroids is catalyzed by phase 1 enzymes whereas the conjugation reactions are catalyzed by phase 2 enzymes. Interest in these reactions exists because the 5α/5β-dihydrosteroids and their tetrahydrosteroids are not always biologically inert, they control steroid hormone metabolism, and inhibitors of these reactions may regulate steroid hormone action. In addition, many synthetic steroids also contain the Δ⁴-3-ketosteroid functionality, and the same enzymes may regulate the duration and efficacy of these drugs which include: steroid contraceptives; anabolic steroids; and inhaled/replacement glucocorticoids. Natural inhibitors of the transformation of these drugs may also cause idiosyncratic drug reactions. This article will focus on the human 3-ketosteroid reductases that are responsible for the formation of the tetrahydrosteroids, and their promiscuity and diversity to act as 17-keto- and 20-keto-steroid reductases that influence the metabolism of androgens, estrogens and progestins.

Fig. 1.

Metabolism of Δ⁴-3-ketosteroids to isomeric tetrahydrosteroids in humans. Enzymes are listed as gene names which are italicized. THS: tetrahydrosteroid.

2. Human ketosteroid reductases (aldo-keto reductases)

Of the human enzymes shown in Fig. 1, all are members of the aldo-keto reductase (AKR) superfamily except SRD5A1–SRD5A3. The AKRs involved include steroid 5β-reductase or AKR1D1[7,8] and the four 3-ketosteroid reductases [9–12]: AKR1C1, 3α(20α)-hydroxysteroid dehydrogenase; AKR1C2, type 3 3α-hydroxysterod dehydrogenase or bile-acid binding protein; AKR1C3, type 2 3α hydroxysteroid dehydrogenase/type 5 17β-hydroxysteroid dehydrogenase; and AKR1C4, type 1 3α-hydroxysteroid dehydrogenase or chlordecone reductase. AKR1C1–AKR1C4 are located on the same chromosome (10p15-p14), they share more than 86% sequence identity, they are available in recombinant form for study, and X-ray crystal structures exist for ternary complexes of AKR1D1 [8,13] and AKR1C1–AKR1C3 [14–17]. Each AKR enzyme adopts the typical (α/β)₈ barrel structure of the AKR superfamily in which an α-helix and β-strand repeats itself eight times so that the β-strands coalesce in the center of the structure to make up the staves of the barrel. At the back of the barrel there are three large loops that help determine substrate specificity/promiscuity. The NADPH cofactor and steroid substrate lie perpendicular to each other so that the nicotinamide head group is in close proximity to the acceptor carbonyl that will be reduced. Steady state kinetic constants k_cat, K_m and k_cat/K_m have been obtained for a large number of steroid substrates [9,11,12], and radiochromatogaphy in combination with liquid chromatography mass spectrometry have identified the products of these reactions [18–21]. Product identification has been an important component of assigning function to these enzymes since if there were reliance solely on changes in absorption or fluorescence of NAD(P)(H), reactions would have been mis-assigned.

Original in vitro characterization of recombinant AKR1C1–AKR1C4 used radiochromatography to identify the reaction products [12]: all enzymes catalyzed the NAD(P)H dependent reduction of [¹⁴C]-5α-dihydrotestosterone (5α-DHT) to yield 5α-androstane-3α,17β-diol (3α-diol); the NAD(P)H dependent reduction of [¹⁴C]-Δ⁴-androstene-3,17-dione to yield testosterone; the NAD(P)H dependent reduction of [¹⁴C]-estrone to yield 17β-estradiol; and the NAD(P)H dependent reduction of [¹⁴C]-progesterone to yield 20α-hydroxyprogesterone. In addition, all enzymes catalyze the corresponding reverse reactions involving the NAD(P)⁺ dependent oxidation of [¹⁴C]-3α-diol to yield 5α-DHT and/or androsterone; the NAD(P)⁺ dependent oxidation of [¹⁴C]-testosterone to yield Δ⁴-androstene-3,17-dione; the NAD(P)⁺ dependent oxidation of [¹⁴C]-17β-estradiol to yield estrone; and the NAD(P)⁺ dependent oxidation of [¹⁴C]-20α-hydroxyprogesterone to yield progesterone. Steady state kinetic parameters indicated that each of these enzymes had 3α/3β-, 17β- and 20α-hydroxysteroid dehydrogenase (HSD) activities to different extents. These studies suggested that AKR1C1 preferred to function as a 20α-HSD, AKR1C2 and AKR1C4 preferred to function as a 3α-HSD; and AKR1C3 preferred to function as a 17β-HSD. However, the function of these enzymes will be dictated by their preference for NADP(H) over NAD(H) and expression levels in hepatic, extrahepatic and steroid hormone target tissues.

3. AKRs function as reductases in vivo

The in vitro characterization of the AKR1C enzymes begs the question as to the directionality of the AKR1C enzymes in vivo. AKR1C enzymes catalyze a sequential ordered bi–bi mechanism in which cofactor binds first and leaves last. Consideration of the K_d values for NADP(H) indicate that this cofactor is bound with nanomolar affinity 9–120 nM while the K_d values for NAD(H) are in the mid-micromolar range >200 µM [22,23]. Thus the enzymes have evolved to use the prevailing concentrations of NAD(P)(H) in the cell and would appear to be bi-directional. However, these values do not consider that low micromolar concentrations of NADP(H) act as potent product inhibitors preventing NAD⁺ dependent oxidation from occurring [24]. In addition, the K_eq and kinetic Haldane show that for the representative rat liver 3α-hydroxysteroid dehydrogenase (AKR1C9) these enzymes favor the reduction reaction [22]. Finally, transient and stable transfection studies of AKR1C cDNA’s into mammalian cells indicate that when faced with using the intracellular concentration of cofactor these enzymes only act as ketosteroid reductases [25–27]. These transfection studies show that AKR1C1 functions as a 20-ketosteroid reductase to inactivate progesterone [27]; AKR1C2 functions as a 3-ketosteroid reductase to inactivate 5α-DHT [24]; AKR1C3 functions as 17-ketosteroid reductase to form the potent steroids testosterone and 17β-estradiol [25,26]; and AKR1C4 functions as a ketosteroid reductase on many 5α- and 5β-dihydrosteroids [12,19]. Thus these enzymes have the capability of regulating ligand concentrations available for the progesterone receptor (PR), androgen receptor (AR) and estrogen receptor (ER), depending on their localization in steroid target tissues. Initial semi-quantitative Northern analysis showed that all four enzymes were expressed in the liver, but that AKR1C4 was liver specific. In addition, it was found that AKR1C3 was highly expressed in the prostate and mammary gland, while AKR1C1, AKR1C2 and AKR1C3 were all expressed in the lung [12].

4. 5α-Dihydrosteroid substrates

As 5α-DHT is the most potent androgen in man the 3-ketosteroid reductase activities of AKR1C1–AKR1C4 were reexamined for their ability to convert 5α-DHT into 3α-diol [20]. Using discriminating radiochromatography and [¹H] NMR it was found that each enzyme produced a mixture of 3α: 3β-diol (5α-androstane-3β,17β-diol) products. However, AKR1C2 and AKR1C4 overwhelmingly favored the formation of 3α-diol with specific activities of 76 and 119 mol/min/mg yielding 3α: 3β-diol ratios of 20 for AKR1C2 and 4.0 for AKR1C4. By contrast AKR1C1 overwhelmingly favored the formation of 3β-diol with a specific activity of 16 nmol/min/mg and a 3α: 3β-diol ratio of 0.25, Fig. 2. This to our knowledge is the first documentation that a HSD can demonstrate mixed stereochemistry or loss of stereo-specificity [20]. Interestingly, AKR1C3 oxidizes 3α-diol to yield 5α-DHT and eventually forms 5α-androstane-3,17-dione via its 3α/17β-HSD activity and also forms epi-androsterone due to its epimerase activity at the 3-keto position. Additionally, AKR1C4 oxidizes 3α-diol to 5α-DHT which is then reduced to 3β-diol demonstrating epimerase activity as well [20].

Fig. 2.

Roles of AKR1C isoforms in the metabolism of testosterone in humans.

The reactions catalyzed by the AKR1C enzymes for the reduction of 5α-dihydroprogesterone (5α-DHP), 5α/5β-dihydroglucocorticoids and 5α/5β-dihydromineralocorticoids still remain to be fully characterized in terms of stereospecificity and kinetic constants. However, AKR1C2 appears to be the major enzyme in peripheral tissues for the reduction of 5α-dihydroprogesterone to form allopregnanlone [28].

5. 5β-Dihydrosteroid substrates

5β-Dihydrosteroids differ from 5α-dihydrosteroids in that they are no longer planar in that their A/B ring cis-ring juncture introduces a 90° bend into the steroid. Knowing that there is variation in the positional and stereospecificity of AKR1C1–AKR1C4, these enzymes were analyzed for their ability to catalyze the reduction of 5β-DHT, 5β-DHP and 20α-hydroxy-5β-pregan-3-one which are the intermediate 5β-reduced metabolites of testosterone and progesterone [19]. Steady state kinetic parameters were assigned and products were identified by LC–MS/MS. AKR1C1–AKR1C4 all reduced 5β-DHT to the corresponding 5β-androstan-3α,17β-diol; where AKR1C4 gave the highest k_cat/K_m value, Fig. 2. AKR1C2 was the only enzyme that produced modest amounts of 5β-androstan-3β,17β-diol. By contrast, with 5β-DHP (5β-pregnane-3,20-dione) AKR1C1–AKR1C4 gave a mixture of products consisting of 3α-hydroxy-5β-pregnane-20-one and 5β-pregnane-3α,20α-diol indicating the presence of both 3-keto and 20-ketosteroid reductase activities. AKR1C1 reduced progesterone to 20α-hydroxyprogesterone and converted 5β-DHP to 20α-hydroxy-5β-pregnan-3-one and 5β-pregnane-3α,20α-diol consistent with its higher 20-ketosteroid reductase activity, Fig. 3.

Fig. 3.

Roles of AKR1C isoforms in the metabolism of progesterone in humans. Reproduced with permission from the Biochemical Society.

6. Metabolism of estrogens

AKR1C1–AKR1C4 will catalyze the NAD(P)H dependent reduction of estrone to 17β-estradiol and the NAD(P)⁺ dependent oxidation of 17β-estradiol to estrone [12]. The assignment of steady state kinetic constants to these reactions proved difficult due to the low turnover number and the difficulty in obtaining reliable constants. AKR1C3 was subsequently found to yield a K_m = 9.0 µM and a k_cat = 0.068 min⁻¹ yielding a k_cat/K_m = 7.6 min⁻¹ mM⁻¹. Focus on this enzyme was predominately due to its 17-ketosteroid reductase activity. Subsequently, AKR1C3 was found to have a significant effect on the conversion of [¹⁴C]-estrone to [¹⁴C]-17β-estradiol when stably transfected into MCF-7 cells. When [¹⁴C]-estrone was incubated with parental MCF-7 cells, substantial endogenous 17-ketosteroid reductase activity was observed. This activity resulted in 53% conversion of 0.1 µM estrone to 17β-estradiol by 24 h. By contrast MCF-7-AKR1C3 cells converted 0.1 µM estrone to 17β-estradiol at a much faster rate. By 6 h, 82% of the estrone had been converted into 17β-estradiol [25]. The importance of the estrone to 17β-estradiol conversion catalyzed by AKR1C3 was supported by the ability of the enzyme to confer a proliferative phenotype on these estrogen-dependent breast cancer cells, and provides an explanation for AKR1C3 overexpression in hormone dependent breast cancer [29–31]. The ability to metabolize estrone suggests that AKR1C enzymes may be involved in the metabolism of synthetic estrogens. However, the majority of these drugs contain a 17α-ethinyl group which precludes metabolism of the 17β-hydroxy group via dehydrogenation.

7. Metabolism of synthetic steroids by AKR1C enzymes

Many synthetic steroids have been developed for therapeutic uses including: the hormone replacement therapeutic tibolone (Livial, [7α,17α]-17-hydroxy-7-methyl-19-norpregn-5(10)-en-20-yn-3-one); the contraceptive steroid norethynodrel [17α-ethynyl-17β-hydroxy-estra-5(10)-en-3-one], and inhaled glucocorticoids e.g., budesonide (Bud; 16,17-(butylidenebis(oxy))-11,21-dihydroxy-(11β,16α)-pregna-1,4-diene-3,20-dione) and flunisolide (Flu; (1S,2S,4R,8S,9S,11S,12S,13R,19S)-19-fluoro-11-hydroxy-8-(2-hydroxyacetyl)-6,6,9,13-tetramethyl-5,7-dioxapentacyclo [10.8.0.0^2,9.0^4,8.0^13,18]icosa-14,17-dien-16-one) for the treatment of asthma, Fig. 4. Target tissue specific effects may be influenced by the metabolism of these steroids by the AKR1C enzymes and their availability for their cognate steroid hormone receptors.

Fig. 4.

Structures of synthetic steroids.

Tibolone is a pro-drug which can be used in the treatment of climacteric complaints and the prevention of osteoporosis but it has no adverse effect on the breast or the endometrium. It exerts its tissue specific effects via tissue-specific metabolism into 3α- and 3β-hydroxytibolone and its Δ⁴-isomer. The 3-hydroxyderivatives inhibit steroid sulfatase and reductive 17β-HSD isoforms that convert estrone to 17β-estradiol in the breast. However, this effect is not seen in osteoblast-like cells in the bone. In addition, the 3α- and 3β-hydroxymetabolites have weak estrogenic activity. Thus, this drug acts as anti-estrogen in the breast but as an estrogen in the bone. By contrast, the Δ⁴-isomer is an agonist for the PR and prevents estrogenic effects on the endometrium [32,33].

AKR1C1, AKR1C3 and AKR1C4 reduced tibolone into products with the same stereochemistry as seen with 5α-DHT as a substrate [21]. Thus, AKR1C1 catalyzed the almost exclusive formation of 3β-hydroxytibolone, AKR1C3 showed weak 3β/3α-HSD activity and AKR1C4 produced only 3α-hydroxytibolone. By contrast, whereas AKR1C2 catalyzed the 3α-reduction of 5α-DHT it catalyzed the 3β-reduction of tibolone. This represented to our knowledge the first example where HSDs invert stereochemistry based on substrate. The preference for AKR1C1 and AKR1C2 to form 3β-hydroxytibolone and the preference for the liver specific AKR1C4 to form 3α-hydroxytibolones explains why 3β-hydroxtytibolone is the major metabolite in target tissues and why 3α-hydroxytibolone is the major circulating metabolite, Fig. 5A. Verification that AKR1C1 and AKR1C2 produce the 3β-hydroxymetabolite came from experiments in HepG2 cells in which AKR1C4 is not expressed [34]. Formation of the 3β-hydroxytibolone was blocked by the nonspecific AKR1C1/1C2 inhibitor flufenamic acid. Using human liver autopsy samples the formation of the 3β- and 3α-hydroxytibolone metabolites was confirmed and the formation of 3α-hydroxytibolone was blocked by the AKR1C4 selective inhibitor phenolphthalein indicating that AKR1C4 was responsible for forming this isomer [34]. Metabolism of norethynodrel which differs from tibolone in that it lacks the 7α-methyl group was found to follow the same stereochemical outcome as observed for tibolone, indicating that the presence of the 7α-methyl group does not influence the stereochemical outcome of the reduction of the 3-ketosteroid by each AKR1C isoform [35].

Fig. 5.

Roles of AKR1C enzymes in tibolone metabolism, top metabolism by discrete AKR1C isoforms in the liver and peripheral tissues; bottom, structures of three AKR1C substrates, 5α-DHT (DHT); tibolone (TIB) and progesterone (PRO). The four rings of the steroids are designated A, B, C and D. The α-face of the steroid is below the plane of the four rings, while the β-face is above the plane of the steroid. The angular methyl groups of DHT (C18 and C19) are β-orientated. The 7-methyl- and 7-ethinyl groups of TIB are α-oriented.

Inhaled glucocorticoids are the mainstay therapy for asthma. However, not all asthmatics respond to these drugs, and even with the best delivery systems a significant portion of the inhaled glucocorticoid ends up in the systemic circulation contributing to side effects of these agents. Thus, it becomes critically important to determine the hepatic and target tissue metabolism of these steroids. Bud and Flu contain a Δ⁴-3-keto group in the A-ring and will not be substrates for the AKR1C enzymes until the A-ring is reduced, Fig. 4. It was found that these inhaled steroids were poor substrates for steroid 5β-reductase (AKR1D1) whereas cortisol is a much preferred substrate [36]. In addition, Bud was a reasonably good inhibitor for AKR1D1 and AKR1C4 which would prolong its systemic half-life and activity whereas this was not the case for Flu. By contrast AKR1C1–AKR1C3 were able to reduce the 20-keto group of both Bud and Flu which could reduce their activity on the lung glucocorticoid receptor (GR) where they are highly expressed. Thus in this instance Bud could prolong its systemic glucocorticoid activity by inhibiting liver specific AKR1D1 and AKR1C4; by contrast both Bud and Flu may have reduced activity in the lung due to the reduction of the 20-keto group by AKR1C1.

8. Metabolism of steroid conjugates

Knowing that AKR1C enzymes have 17-keto- and 20-ketosteroid reductase activity raised the issue as to whether other 17β-substituents might be tolerated by these enzymes. Examination of the crystal structures of these enzymes showed that the large loop structures do not fold around or cap the C17-position suggesting that this open space can accommodate bulk at this position. This led us to examine the turnover of 5α-DHT-17β-glucuronide, 5α-DHT-17β-sulfate and tibolone 17β-sulfate by the AKR1C enzymes [18]. Products were characterized by LC–MS and kinetic parameters assigned. The product profile observed with the 5α-DHT conjugates was similar to that observed when unconjugated 5α-DHT was used as substrate. The exception was that AKR1C2 inverted its stereochemistry so that it performed 3β-reduction with 5α-DHT-17β-glucuronide while performing 3α-reduction with 5α-DHT and 5α-DHT-17β-sulfate. This inversion of stereochemistry observed with AKR1C2 is reminiscent of what was observed during the 3-keto reduction of 5β-DHT, tibolone and norethynodrel [21,35]. The catalytic efficiency of the enzymes was significantly impaired by the presence of the 17β-glucuronide but not by the 17β-sulfate group. The impairment in catalytic efficiency was not due to a change in stereochemical outcome since both 5α-DHT-17β-glucuronide and tibolone 17β-sulfate are 3β-reduced by AKR1C2 but the latter conjugate is reduced with a catalytic efficiency that is 40-times greater. AKR1C4 was the enzyme that displayed superior catalytic efficiencies versus the other isoforms for both the 17β-glucuronide and 17β-sulfate conjugates. This raises the prospect that functionalization of the A-ring of steroids such as 5α-DHT does not have to occur before conjugation reactions proceed for 17β-hydroxysteroids. For these steroids phase 2 conjugation reactions may precede phase 1 functionalization reactions in the liver [18].

9. Structural basis for steroid promiscuity

Steroid hormone transforming AKRs were originally described as HSDs. One of the original tenets associated with this enzyme group were that they were positional and stereospecific for the reactions that they catalyze [37]. Neither of these tenets hold for the AKR1C enzymes. All of the enzymes are promiscuous with regards to their positional specificity but show distinct preferences. In addition, the AKR1C enzymes also show different modes of substrate binding and/or inversion of stereochemistry based on the steroid substrate utilized. This raises the issue as to the structural basis for this promiscuity or diversity in substrates. Both enzyme structure and substrate recognition seem to play a role.

AKR1C1 acts preferentially as a 20-ketosteroid reductase and AKR1C2 acts preferentially as a 3-ketosteroid reductase forming either 20α- or 3α-reduced products, respectively, suggesting that the steroid substrate needs to bind backwards in AKR1C1 versus the position seen in AKR1C2, i.e., the D-ring must bind in the A-ring position at the active site. AKR1C1 and AKR1C2 have 97% sequence identity and the major structural difference between the two enzymes is in only seven amino acids, and only one difference occurs at the active site where Leu54 in AKR1C1 is replaced by the smaller Val54 in AKR1C2. Mutagenesis studies on AKR1C1 show that the L54V mutant is sufficient to change the ability of AKR1C1 to act as a 20-ketosteroid reductase to a 3-ketosteroid reductase [38]. The reverse mutation converts AKR1C2–AKR1C1 [39]. In these enzymes, the stereochemistry of hydride transfer is rigid and 4-pro-R-hydride transfer must occur to produce the desired product.

Insight into why these changes alter substrate specificity come from crystal structures of the AKR1C1-NADP⁺-20α-hydroxyprogesterone ternary complex (PDB 1MRQ), the AKR1C2-NADP⁺-ursodeoxycholate ternary complex (PDB 1IHI), and the AKR1C2 V45L-NADP⁺-progesterone ternary complex (PDB 4L1X) [14,15,39]. These structures show that as predicted in AKR1C1 the D-ring of the steroid binds at the base of the active site so that the 20-keto group would lie in the oxyanion hole. The C18 and C19 angular methyl groups point away from the bulky Leu54 and encounter Trp227 and Leu308 on the opposite side of the steroid cavity so progesterone would be the preferred substrate for AKR1C1.

In the AKR1C2-NADP⁺-ursodeoxycholate ternary complex, the C24 carboxylic acid binds in the oxyanion hole at the active site, but the steroid is now bound backwards and upside down relative to the position of testosterone in the AKR1C9-NADP⁺-testosterone ternary complex (PDB 1AFS) where the 3-keto group of the competitive inhibitor is in the oxyanion hole [40]. Thus, the C1 8 and C19 angular methyl groups of ursodeoxycholate are inverted and now face Val54 in AKR1C2, Fig. 6.

Fig. 6.

Different binding modes for ursodeoxycholate and testosterone in AKR1C2 (green) and in AKR1C9 (rat liver 3α-hydroxysteroid dehydrogenase (blue). Reproduced with permission from the American Chemical Society.

This structure provides a glimpse of different binding modes that could be used to produce 3α-, 17α- and 20α-hydroxysteroids. Examination of the crystal structure of AKR1C2-NADP⁺-progesterone (PDB 4L1W) and AKR1C2V54L-NADP⁺-progesterone is more revealing. In the former structure progesterone can assume two binding modes one in which the steroid binds in the same manner as ursodeoxycholate in AKR1C2 and another in which progesterone binds in the same manner as testosterone in AKR1C9. Thus the steroid substrate can flip over. However, the AKR1C2V54L-NADP⁺-progesterone structure only shows one binding mode for the steroid in which the C18 and C19 angular methyl groups point away from Leu54 due to the narrowly restricted steroid binding cavity, i.e., it assumes the binding pose of 20α-hydroxyprogesterone in the AKR1C1 structure. Thus, reduction of progesterone is accomplished in AKR1C1 due to backwards binding of the steroid and steric forces that prevent the C18 and C19 methyl groups interacting with Leu54.

AKR1C1 conducts 3β-reduction on 5α-DHT whereas AKR1C2 conducts 3α-reduction on the same substrate; and both perform 3β-reduction on tibolone. To further understand the basis for this selectivity we performed molecular modeling simulations using Autodock where the validity of the docking procedure was confirmed by showing that 20α-hydroxprogesterone and ursodeoxycholate could be docked in silico into the same position observed in the respective crystal structures of AKR1C1 and AKR1C2 [41]. These studies revealed that when 5α-DHT binds to AKR1C1 the A-ring of the steroid binds in the oxyanion hole but the bulky Leu54 pushes the β-face of the steroid across the active site to make contact with Trp227, thus the α-face is presented to the nicotinamide head group of the cofactor resulting in the formation of 3β-diol. In this binding pose the C18 and C19 angular methyl groups point away from Leu54 as seen for the binding of progesterone to the AKR1C2V54L mutant. When 5α-DHT binds to AKR1C2 the A-ring of the steroid binds in the oxyanion hole but the α-face of the steroid hugs the side of the channel containing Val54 so that the β-face of the steroid is presented to the nicotinamide head group of the cofactor resulting in the formation of 3α-diol, Fig. 7. In this binding pose the C18 and C19 angular methyl groups still point away from residue 54. Thus, two possibilities exist to explain the inversion of stereochemistry observed with 5α-DHT as substrate for AKR1C1 and AKR1C2. In one instance the steroid is pushed across the active site as suggested by Autodock; while in the other instance the steroid is flipped over as shown by the crystal structures of AKR1C2-NADP⁺ursodeoxycholate and the AKR1C2-NADP⁺-progesterone complexes. Further X-ray crystal structures will help to distinguish between these two possibilities. Thus, this one amino acid substitution V54L is sufficient to permit steroid substrates to bind backwards D-ring in the A-ring position to change positional specificity; and it is sufficient to change stereochemistry of hydride transfer to the 3-ketone group while the position of the cofactor remains invariant.

Fig. 7.

Inversion of stereochemistry in the 3α- to 3β-reduction of ketosteroids. Docking of 5α-DHT into AKR1C1 (light grey) showing formation of a 3β-reduced product; and docking of 5α-DHT into AKR1C2 (black) showing formation of a 3α-reduced product.

The ability of AKR1C2 to catalyze the 3β-reduction of tibolone and norethynodrel when this enzyme prefers to catalyze the 3α-reduction of 5α-DHT also requires an explanation. In tibolone, the 7-methyl and 17-ethinyl groups are α-oriented, and in noerthynodrel the 17-ethinyl group is α-oriented. This α-orientation of the substituents makes the α-face of the steroid mimic the β-face of the steroid with its two angular groups, Fig. 5B. Thus it is predicted that these synthetic steroids bind upside down relative to 5α-DHT i.e., angular methyl groups inverted in the AKR1C2 active site [35,41].

AKR1C3 acts preferentially as 17-ketosteroid reductase to produce 17β-reduced products. Inspection of the crystal structure of the AKR1C3-NADP⁺. Δ⁴-androstene-3,17-dione (PDB 1XF0) shows that position of the 17-keto group is such that it approaches the oxyanion hole as expected i.e., the D-ring is in the A-ring position [17]. However, the C17-ketone group is 7 A° away from the nicotinanide head group of the cofactor so the steroid is not in a productive binding mode. AKR1C3 like AKR1C1 contains Leu54. Stereochemistry in favor of a 17β-product is likely explained if the β-face of the steroid is pushed by steric forces towards Trp227 and as a result the α-face of the steroid is presented to the nicotinamide head group to produce the 17β-product. Currently, no crystal structure exists for AKR1C4 in complex with a steroid so that it is not possible to use structural information to explain its stereo-specificity. Like AKR1C1 and AKR1C3, it contains Leu54 yet it reduces 3-ketosteroids to the 3α-product. Thus, other structural constraints are in place to produce these stereospecific products.

10. Pre-receptor regulation of nuclear receptor action

AKR1C1, AKR1C2 and AKR1C3 are poised to regulate the concentration of ligands available for nuclear receptors based on their preferred 20-, 3-, and 17-ketosteroid reductase activities. AKR1C1 by converting progesterone (active hormone) to 20α-hydroxprogesterone (inactive hormone) can regulate occupancy of the PR. Expression of AKR1C1, in the breast, mammary gland, endometrium, endometrial epithelial cells as well as transient transfection studies supports this assertion [12,27,42]. As described above, AKR1C1 may also regulate the occupancy of GR by synthetic glucocorticoids; and expression of AKR1C1 in human lung and human lung cells supports this notion [43]. Interestingly, AKR1C1 also converts 5α-DHT to 3β-diol a proapotopic ligand for ERβ and suggests that it may play an important anti-proliferative role in tissues and cells that express ERβ [20]. Notably, AKR1C1 is found in the breast and prostate; and prostate stromal and epithelial cells [44].

AKR1C2 is the primary peripheral enzyme involved in the inactivation of 5α-DHT (K_d = 10⁻¹¹ M for the AR) to 3α-diol (K_d = 10⁻⁶ M for the AR) and deprives the AR of its ligand [45]. This is supported by transient transfection studies in COS-1 and PC-3 cells [24], expression profiling in prostate stromal and epithelial cells [44]; and expression analysis in prostate biopsy material. AKR1C2 is also expressed in the CNS where it is the primary enzyme for the conversion of 5α-DHP to allopregnanolone [28] which is a neuroactive steroid and functions as an allosteric effector of the GABAa receptor, to promote chloride channel opening and inhibitory potentials.

AKR1C3 is an important peripheral 17-ketosteroid and regulates the level of ligands available for the AR [46,47]. It converts Δ⁴-androstene-3,17-dione (weak androgen) to testosterone (potent androgen); and 5α-androstane-3,17-dione (weak androgen) to 5α-DHT (potent androgen). It is upregulated in castration resistant prostate cancer where it contributes to intraprostatic androgen biosynthesis and a number of inhibitor programs exist for targeting this enzyme [48,49]. Its activity has been confirmed by stable transfection studies in LNCaP cells and by overexpression and knock -down in VCaP cells [26,50]. AKR1C3 also converts estrone (weak estrogen) to 17β-estradiol (potent estrogen) [12]; and its ability to make testosterone in peripheral tissues also provides a substrate for aromatase to make 17β-estradiol. Through these combined activities, it can regulate occupancy of the ER in the breast. This has been confirmed by stable transfection in MCF-7 cells [25] and by its overexpression in ductal carcinoma in situ [30].

11. Summary

Human AKR1C enzymes are pluripotent enzymes which are ketosteroid reductases that may have a profound effect on steroid hormone metabolism in the liver and steroid hormone action at the pre-receptor level in target tissues. They may also determine the efficacy of synthetic steroids and their ability to bind to their cognate receptors. Fundamental enzymology and product characterization is still required to annotate the roles of AKR1C1–AKR1C4 in the metabolism of 5α-dihydroprogesterone, 5α/5β-dihydroglucocorticoids and 5α/5β-dihydromineralocorticoids. With the uniform adoption of real-time PCR and microarray expression studies, changes in transcript levels in different microenvironments will likely reveal more details about how the expression of these enzymes is altered and how they are regulated in an autocrine and paracrine fashion. The AKR1C genes contain many steroid response elements suggesting that they are under steroid hormone control [51]. Those enzymes involved in the inactivation of steroid hormones may well be upregulated by hormone excess and down regulated by hormone deprivation. The reverse may be true for those enzymes that make active steroid hormones e.g., AKR1C3. This raises the issue that unless attention is focused on the levels of steroid hormones present in cell culture e.g., FBS containing media, or charcoal dextran stripped media the level of expression of these enzymes may change in the course of an experiment. The function of the AKR1C enzymes can also be further revealed by si-RNA/sh-RNA approaches and the use of isoform specific inhibitors as chemical probes.