The hop-derived compounds xanthohumol, isoxanthohumol and 8-prenylnaringenin are tight-binding inhibitors of human aldo-keto reductases 1B1 and 1B10

Abstract Xanthohumol (XN), a prenylated chalcone unique to hops (Humulus lupulus) and two derived prenylflavanones, isoxanthohumol (IX) and 8-prenylnaringenin (8-PN) gained increasing attention as potential anti-diabetic and cancer preventive compounds. Two enzymes of the aldo-keto reductase (AKR) superfamily are notable pharmacological targets in cancer therapy (AKR1B10) and in the treatment of diabetic complications (AKR1B1). Our results show that XN, IX and 8-PN are potent uncompetitive, tight-binding inhibitors of human aldose reductase AKR1B1 (Ki = 15.08 μM, 0.34 μM, 0.71 μM) and of human AKR1B10 (Ki = 20.11 μM, 2.25 μM, 1.95 μM). The activity of the related enzyme AKR1A1 was left unaffected by all three compounds. This is the first time these three substances have been tested on AKRs. The results of this study may provide a basis for further quantitative structure–activity relationship models and promising scaffolds for future anti-diabetic or carcinopreventive drugs.


Introduction
The female inflorescences of hops (Humulus lupulus) have long been used in traditional Chinese and Ayurvedic medicine mainly to treat sleep disturbances 1 and are also known for their antibiotic properties 2 . Among the secondary plant compounds that occur in the resinous inflorescences of H. lupulus are prenylated chalcones and other flavonoids 3 . These compounds are well known for their bittering and preserving qualities in the brewing process of beer 3,4 , as well as for their bioactivity (antibiotic, anti-viral and antioxidant properties) [5][6][7] . Recently, the prenylated chalcone xanthohumol (XN) has come into focus of biomedical research, due to its versatile bioactive characteristics 8 . Since XN is a unique prenylflavonoid occurring in hops, beer is the only noteworthy dietary source for XN in central Europe. In addition to XN, hops inflorescences also contain flavanones like isoxanthohumol (IX) and 8-prenylnaringenin (8-PN) 9 ( Figure 1), but at 10-to 100-fold lower concentrations than XN. The main metabolic routes for XN, IX and 8-PN have been characterised extensively, both, in vitro by using human and rat liver microsomes 3,10-12 and in vivo, in rat models 13,14 . XN is subject to spontaneous conversion into IX via intramolecular Michael addition 15 . IX, in contrast, undergoes an enzymatic O-demethylation via hepatic CYP1A2 to 8-PN, which has been described as the most potent phytoestrogen found in nature 5,16 . All three substances are currently under investigation due to their anti-diabetic 17,18 , anti-carcinogenic 3,8,19 and antioxidant properties 20 . Moreover, XN has been shown to have anti-HIV traits 21,22 .
The pathogenesis of several inflammatory conditions, such as diabetes, cancer and sepsis, have in many cases been linked to enzymes of the aldo-keto reductase (AKR) superfamily [23][24][25][26][27][28][29] . Among them, AKR1A1, also known as human aldehyde reductase, is involved in the reduction of biogenic and xenobiotic carbonyl group containing compounds, such as the cytotoxic lipid peroxidation-derived aldehydes 4-ONE and 4-HNE 30,31 and the chemotherapeutic drugs doxorubicin (DOX) or daunorubicin (DAUN) 32 . It is widely expressed in most tissues, with particularly high levels reported in the cerebrum, liver and kidneys 33 .
The same applies for the homologous enzyme AKR1B1 (60% sequence identity to AKR1A1), commonly known as human aldose reductase 34 . AKR1B1 mediates the first step of the "polyol pathway" by reducing glucose to sorbitol through a NADPHdependent reaction. Further, sorbitol is oxidised to fructose by sorbitol dehydrogenase, using NAD þ as a co-factor 35,36 . Under normoglycemic conditions only 3% of total glucose is converted into sorbitol through the polyol pathway. Under hyperglycemic conditions, however, the flux of glucose through this metabolic pathway may increase up to 10-fold (30%), which might finally lead to accumulation of excess sorbitol 35,[37][38][39] . The resulting osmotic stress and imbalances of the pyridine nucleotide redox status decrease the cell's antioxidative capabilities. In diabetes mellitus, this promotes the formation of advanced glycation end products (AGEs) 34,40 , which themselves lead to diabetic complications, such as microangiopathies, nephropathies, retinopathies, peripheral neuropathies and cataract [41][42][43] . Consequently, aldose reductase inhibitors (ARIs) are at the focus of exploratory pharmaceutical research, as they yield the potential to prevent or control the onset of these diabetic complications. In vitro, AKR1B1 displays poor affinity to glucose (K M ¼ 50-100 mM) and hydrophobicity of the putative substrate-binding domain essentially precludes efficient carbohydrate reduction 44 . Therefore, lipid peroxidation-derived hydrophobic aldehydes, such as 4-hydroxynonenal (HNE) and other related nephro-and hepatotoxic compounds are more likely to be physiological substrates of AKR1B1. Finally, high catalytic efficiencies of AKR1B1 towards these compounds suggest an important metabolic role in the detoxification of lipid-derived aldehydes 27,44,45 . Physiologically and despite the relatively high K M for glucose, however, selective AKR1B1 inhibition remains an effective treatment option for diabetic complications derived from glucose-related adducts 44,46 .
Overlapping substrate spectra are considered a major pitfall of many ARIs, as they often inhibit AKR1A1 and AKR1B1 simultaneously, thereby interfering with their central role in detoxification processes. This has raised the demand for selective inhibitors that preferentially bind to AKR1B1 and leave the activity of AKR1A1 unaffected [47][48][49] .
Additionally, the AKR1B subclass contains the AKR1B1 homologue AKR1B10 (71% sequence identity to AKR1B1), a NADPHdependent oxidoreductase that converts both endo-and exogenous carbonyl group containing compounds to their corresponding alcohols [50][51][52] . In contrast to AKR1B1, AKR1B10 exhibits a more restricted substrate specificity towards isoprenyl aldehydes including farnesal and geranylgeranial, which are reduced to their corresponding alcohol metabolites 52 .
These metabolites are involved in protein prenylation, a process suspected of being crucial for carcinogenesis 28 . Moreover, AKR1B10 participates in the reduction of retinal to retinol, and thereby balances the homeostasis of retinoic acid, which is able to modulate cell proliferation and differentiation 53,54 , as well as the reductive metabolism of carbonyl group containing exogenous compounds [50][51][52] .
AKR1B10 is physiologically expressed in the small and large intestine and in the adrenal glands 55,56 . AKR1B10 overexpression has been reported in several types of cancer, including lung 57,58 , pancreatic 26,28,59 and hepatocellular tumors 24 . At the same time, AKR1B10 fulfills a regulatory role in cell proliferation and differentiation by modulating the metabolism of retinoids and prenylation of oncoproteins, including Kras 26,59 . Since AKR1B10 overexpression is already detectable in precancerous lesions 50,60 , metabolic events mediated by this enzyme could play a crucial role in the development of cancer. In particular, membrane proteins like Ras and Rasrelated GTP-binding proteins rely on prenylation to exert their functions in processes of cellular growth and differentiation 28,61,62 . Within the metabolism of farnesal and geranylgeranial whose reduced metabolites are important intermediates of cholesterol synthesis and protein prenylation, AKR1B10 performs key reactions by reducing farnesyl and geranylgeranial to farnesol and geranylgeraniol. The latter are further phosphorylated to farnesyl and geranylgeranyl pyrophosphates, two main intermediates of cholesterol synthesis involved in protein prenylation 28 . Increased levels of AKR1B10 expression in neoplastic cells alterate prenylation processes and post-translational modifications of the aforementioned protooncoproteins and, hence, accelerate tumor formation 28,53 .
Thus, the finding of new potent and selective inhibitors for both AKR1B10 and AKR1B1 are current challenges in biomedical research, in order to prevent uncontrolled cell proliferation and to reduce microangiopathies, as found in diabetic complications. In this article, we investigated the inhibitory potential of three hopderived substances on three members of the AKR superfamily (namely AKR1A1, AKR1B1 and AKR1B10). Our results provide evidence that XN, IX and 8-PN are potent inhibitors of AKR1B1 and AKR1B10, while not affecting the activity of the closely related enzyme AKR1A1.

Preparation of recombinant proteins
The recombinant enzymes AKR1A1, AKR1B1 and AKR1B10 were prepared in an Escherichia coli expression system according to previously published methods: AKR1A1 was a gentle gift from Prof. Dr. Vladimir Wsol 63 ; AKR1B1 was a friendly gift from Dr. Nina Kassner; information about production and purification of AKR1B10 has been published before 64 . Genetic information on the specific inserts of all obtained plasmids was verified by sequencing (MWG Eurofins). The plasmids were then transformed into E. coli BL21 (DE3) cells. For over-expression of 6 Â His-tagged enzymes, a 400-ml culture (containing the appropriate antibiotic; plasmid dependent) was grown to an optical density of 0.6 at 600 nm at 37 C. Protein over-expression was induced by adding isopropyl-1-thio-galactopyranoside (IPTG) to the culture medium (final concentration of 1 mM). After 3 h, cells were harvested by centrifugation (6000 g, 15 min) and re-suspended in 20 ml PBS-I buffer (20 mM NaH 2 PO 4 , 500 mM NaCl, 10 mM imidazole, pH 7.4). Cell disruption was performed by ultrasonication with cooling on ice, to avoid heating. The sample was subsequently centrifuged at 100,000 g at 4 C for 1 h. The obtained supernatants, containing the target protein were purified using Ni-affinity chromatography on an € AKTA-Purifier System (Amersham Pharmacia Biotech, Uppsala, Sweden). Purification progress was monitored by SDS-PAGE of the obtained fractions (not shown). Enzyme concentrations were determined using a Qubit 2.0 fluorometric quantitation system (Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions.

Determination of inhibition parameters
Catalytic properties were determined by measuring the decrease in absorbance at 340 nm at 37 C (Cary 100 scan photometer, Varian, Pal Alto, CA, USA). A reaction mixture without inhibitor consisted of different concentrations of substrate (see Table 1 for details), 200 lM NADPH, 0.1 M NaH 2 PO 4 buffer (pH 7.4) and an appropriate amount of enzyme in a total assay volume of 0.8 ml. Final enzyme concentrations in the assay ranged from 583 nM (AKR1B10) to 712 nM (AKR1B1). For inhibitor selectivity studies on AKR1A1, AKR1B1 and AKR1B10 stock solutions of the inhibitors XN, XI and 8-PN were prepared in dimethyl sulfoxide (DMSO). The final concentration of DMSO in the assay was 0.5%. Activity measurements were started without pre-incubation by adding an appropriate amount of enzyme. When collecting data for dose-response curves, initial velocities of the glyceraldehyde reduction (concentration at K M ; enzyme specific) in the presence of inhibitors were assayed as described above. The percentage of inhibition was calculated considering the activity in the absence of inhibitor to be 100%.
Initially, the half maximal inhibitory concentrations (IC 50 values) were determined for each inhibitor in the presence of each enzyme, using the shared substrate glyceraldehyde (set to their specific Km; 3.6 mM, 50 mM and 4 mM for AKR1A1, AKR1B1 and AKR1B10, respectively) to assess specificity among the structurally similar members of the AKR-superfamily. For IC 50 determination, experimental data were normalised and fitted to a sigmoidal curve as implemented in GraphPad6 (GraphPad Software Inc., La Jolla, CA, USA). Whenever tight-binding inhibition was observed, the inhibition constant K i was determined by fitting inhibition data to the Morrison equation as implemented in GraphPad Prism6 (GraphPad Software Inc., La Jolla, CA, USA) 65 , using non-linear regression.
In order to verify the inhibitory potency, enzyme-specific physiological substrates for AKR1B1 (glucose, K M ¼ 32 mM) and AKR1B10 (farnesal; K M ¼ 5 mM) were used to determine inhibition parameters. Enzyme inhibition parameters were assayed as described above. The inhibition mechanism of each compound for the respective enzymes was analysed by plotting IC 50 values at different substrate concentrations (at least five inhibitor and substrate concentrations) 65,66 . All data obtained were plotted and analysed using GraphPad Prism6 (GraphPad Software Inc., La Jolla, CA, USA).

Determination of inhibitor selectivity
Initially, dose-response curves for XN, IX and 8-PN with AKR1A1, AKR1B10 and AKR1B1, using glyceraldehyde, were calculated (IC 50and K i -values are summarised in Table 1). Figure 2 exemplarily shows the determination of IC 50 -and K i -values for IX with AKR1B1. IX turned out to be the most effective inhibitor among the three substances for both AKR1B1 and AKR1B10 (IC 50 ¼ 0.57 and 1.09 mM, respectively). The IC 50 for IX is 6 to 15 times lower than compared to XN (Table 1). Interestingly, the activity of AKR1A1 was unaffected by all three substances (IC 50 > 50 mM).

Determination of inhibition parameters for physiological substrates
Inhibition constants, as well as the mode of inhibition on the reduction of two specific physiological substrates, were further calculated by using glucose for AKR1B1 and farnesal for AKR1B10, at concentrations equal to their corresponding K M values (as determined before: 32 mM and 5 mM, respectively-data not shown). Figures 3 and 4 show the exemplified determination of the IC 50and K i -values (calculated using the Morrison equation 65 ), in presence of IX, for AKR1B1 and AKR1B10, respectively. Tables 2 and 3 summarise the obtained IC 50 and the inhibition constants for the respective enzymes.  Since all three substances turned out to be tight-binding inhibitors of AKR1B1 and AKR1B10 (ratio [Inhibitor]:[Enzyme] ranging from 0.9 to 8.2), IC 50 -values were determined at five different substrate concentrations, including K M to analyse the mode of inhibition. All inhibitors exhibited an uncompetitive mode of inhibition for both, AKR1B1 and AKR1B10, in the presence of their physiological substrates (Figures 5 and 6).
A closer examination of the binding mechanism of the three inhibitors was carried out by investigating their mode with respect to the co-factor binding site for NADPH. Interestingly, all the three substances behave as uncompetitive inhibitors of AKR1B10, while only IX displays the typical pattern of an uncompetitive inhibitor on AKR1B1, whereas XN and 8-PN inhibit in a non-competitive fashion (see Figure 7).

Discussion
Aldose reductase (AKR1B1) and AKR1B10 are involved in numerous pathologies and have therefore been proposed as suitable targets     for drug development. Under hyperglycemic conditions, AKR1B1 promotes osmotic imbalance and modifies physiological ratios of the redox couples of both NAD þ and NADP þ , with decreasing antioxidant defense abilities. Finally, it induces a potential increase in protein glycation phenomena due to an increase in AGE levels. This has led to the identification of AKR1B1 as a primary target to prevent the onset of secondary diabetic effects. AKR1B1's homologue AKR1B10 is overexpressed in multiple cancer types, including malignancies of breast, prostate and lungs, and is deemed suitable as a target in cancer treatment. Recent studies have shown that plant/hops specific XN, IX and 8-PN act as anti-diabetic, anti-carcinogenic and antioxidative agents. Hitherto, it has not been explored how this mode of action relates to the role of AKRs in the pathogenesis of the complications described above.
Our results show that XN, IX and 8-PN are potent inhibitors of the AKR1B's subfamily members B1 and B10 with IC 50 and K i values ranging at low mM or nM concentrations, using D,L-glyceraldehyde as a generic substrate (see Table 1). In the following, glucose (AKR1B1) and farnesal (AKR1B10) were used as corresponding physiological substrates (see Table 2). After all, IX turned out to be the most potent inhibitor on both AKR1B1 and AKR1B10, with all substrates tested (K i ¼ 0.17 lM and 0.52 lM with AKR1B1 and AKR1B10 in presence of GA as substrate and K i ¼ 0.34 lM and 2.25 lM with AKR1B1 and AKR1B10 in presence of glucose and farnesal, respectively).
Interestingly, all three inhibitors exhibit a 4 0 -OH group and a 2-benzyl substituent in their structure (see Figure 1). In a previous work, Rastelli et al. 67 have shown that a 4 0 -OH group is crucial for the inhibition of AKR1B1 by 7-hydroxy-2-(4 0 -hydroxybenzil)-4H-1benzopyran-4-one, a modified form of the potent ARI quercetin, and 2 0 ,4,4 0 -trihydroxychalcone, and that it specifically binds to Thr113. The same work further showed that the 2-benzyl substituent, owing to its hydrophobic aromatic nature and conformation, optimally fits an additional hydrophobic pocket of the enzyme, lined by Trp111 and Leu300. In particular, the binding of this substituent to this additional pocket provides a selective kind of inhibition for AKR1B1 with respect to the closely related enzyme AKR1A1 67 .
There is also an evidence of prenylated flavonoids, that possess additional hydrophobic and anionic characteristic moieties (prenyl groups) at the C-8 position of their flavonoid skeletons, playing important roles in aldose reductase inhibition 68,69 . In this sense, a comparison can be made between 8-PN (AKR1B1 IC 50 ¼ 0.81 lM obtained in this study) and its isomer 6-prenylnaringenin (6-PN), which is also a phytoestrogenic prenylflavanone occurring in hops (AKR1B1 IC 50 ¼ 6.2 lM 70 ). In fact, 8-PN exhibits an inhibitory potency that is 7.6 times greater as compared to that of 6-PN, applying glyceraldehyde as substrate. This might as well stress the importance of prenyl moieties at the C-8 position in terms of AKR1B1 inhibition. Although no IC 50 data for 6-PN in presence of AKR1B10 were available at the time of this study, the low IC 50 value obtained for 8-PN, might as well indicate a similar effect.
We further demonstrated that the three substances have no inhibitory effect on the closely related AKR member AKR1A1. This is an important result since many candidate inhibitors failed in clinical trials due to their undesired property of inhibiting both enzymes, AKR1B1 and AKR1A1 25,44 . Nevertheless, none of the inhibitors characterised in this study showed a selective inhibition, neither for AKR1B1 nor for AKR1B10. However, our studies may provide a basis for further modification of the molecules' scaffolds, in order to design novel selective inhibitors for the respective enzymes. In fact, plenty of current studies aimed to find selective inhibitors for AKR1B1 over AKR1B10, and vice versa, in order to exclusively treat complications of diabetes or cancer. Yet, only a few selective inhibitors have been identified so far, including Androst-3b,5a,6b,19-tetrol (IC 50 ¼ 0.86 mM) or oleanolic acid (IC 50 ¼ 0.09 mM) [71][72][73][74][75] . So far, numerous tested phytogenic inhibitors have shown only moderate degrees of selectivity 71,76 .
Our results show that IX (IC 50 ¼ 0.57 lM for AKR1B1 and IC 50 ¼ 1.09 lM for AKR1B10) and 8-PN (IC 50 ¼ 0.81 lM for AKR1B1 and IC 50 ¼ 0.99 lM for AKR1B10) inhibit both enzymes with an efficacy 6 to 15 times greater than XN when applying GA as substrate (Table 1).
In presence of the physiological substrates, glucose and farnesal, the efficacy of IX (IC 50 ¼ 0.88 lM for AKR1B1 and IC 50 ¼ 0.63 lM for AKR1B10) and 8-PN (IC 50 ¼ 1.87 lM for AKR1B1 and IC 50 ¼ 3.96 lM for AKR1B10) is 10-30 times greater than that of XN (Tables 2 and 3).  IX and 8-PN are strongly related prenylated flavonoids, whereas XN is a prenylated chalcone that exhibits an open ring in its core structure, which might serve as a possible explanation for its lower inhibitory potency. Our enzyme inhibition experiments showed that all compounds under investigation exhibit an uncompetitive mode of inhibition on both, AKR1B1 and AKR1B10. Uncompetitive inhibitors exclusively bind to the enzyme-substrate complex (ES), thus circumventing competition with the substrate for the active site. For AKR1B1, this constitutes a promising mechanism of inhibition, since the affinity of uncompetitive inhibitors is greatest at saturating concentrations of the substrate, as observed under hyperglycemic conditions (>7 mM, diabetic range) 74,77 . Theoretically, under the aforementioned pathological conditions, a drastic inhibition of AKR1B1 would occur even at low inhibitor concentrations.
In order to closely examine the binding mechanism of the substances, further experiments were performed to investigate whether XN, IX and 8-PN compete with NADPH at the co-factor binding site. However, our work revealed no competition between the three tested substances and NADPH, neither for AKR1B1, nor for AKR1B10. All substances tested exhibited an uncompetitive mode of inhibition for AKR1B10 with respect to NADPH, while only IX showed an uncompetitive inhibition pattern for AKR1B1, with respect to the co-factor binding site. XN and 8-PN exhibited a non-competitive mode of inhibition on AKR1B1.
Interestingly, the abovementioned substances show an uncompetitive mode of action with respect to both, the substrate and the co-factor binding site. Hypothetically speaking, this would mean that the inhibitor does not bind to the enzyme-NADPH binary complex as expected, but, instead, to the subsequent and transient enzyme conformation after the product release but before the release of NADPþ. This mechanism was previously hypothesised by Harris and Kozarich, 1997 78 and Copeland 2005 79 for the binding of epristeride to the NADPH-dependent steroid-5a reductase (M. mulatta).
Hence, XN, 8-PN, IX for AKR1B10, and IX for AKR1B1 might act as uncompetitive inhibitors that bind to an enzyme species, following the formation of the initial ES complex. More challenging to explain is the behavior of 8-PN and XN, which both have been showing a non-competitive mode of inhibition with respect to AKR1B1 at the NADPH binding site.
Since the models obtained from molecular docking experiments (data not shown) did not clarify the binding mechanism of the three compounds, further investigations by X-ray diffraction are required in future studies.
Overall, we could show that the hop compounds tested have a selective inhibitory effect on the AKR1B subfamily, with respect to AKR1A1. All substances exhibit an uncompetitive mode of inhibition on both, AKR1B1 and AKR1B10. This fact might be promising when it comes to designing novel drugs and molecular therapies with high efficacy even at low concentrations. The results of this study may also provide a basis for further quantitative structure-activity relationship models and favorable scaffolds for new selective inhibitors of AKR1B1 or AKR1B10.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
Authors acknowledge financial support by Land Schleswig-Holstein within the funding program Open Access Publikationsfonds.