Selective Inhibition of Human AKR1B10 by n-Humulone, Adhumulone and Cohumulone Isolated from Humulus lupulus Extract

Hop-derived compounds have been subjected to numerous biomedical studies investigating their impact on a wide range of pathologies. Isomerised bitter acids (isoadhumulone, isocohumulone and isohumulone) from hops, used in the brewing process of beer, are known to inhibit members of the aldo-keto-reductase superfamily. Aldo-keto-reductase 1B10 (AKR1B10) is upregulated in various types of cancer and has been reported to promote carcinogenesis. Inhibition of AKR1B10 appears to be an attractive means to specifically treat RAS-dependent malignancies. However, the closely related reductases AKR1A1 and AKR1B1, which fulfil important roles in the detoxification of endogenous and xenobiotic carbonyl compounds oftentimes crossreact with inhibitors designed to target AKR1B10. Accordingly, there is an ongoing search for selective AKR1B10 inhibitors that do not interact with endogeneous AKR1A1 and AKR1B1-driven detoxification systems. In this study, unisomerised α-acids (adhumulone, cohumulone and n-humulone) were separated and tested for their inhibitory potential on AKR1A1, AKR1B1 and AKR1B10. Also AKR1B10-mediated farnesal reduction was effectively inhibited by α-acid congeners with Ki-values ranging from 16.79 ± 1.33 µM (adhumulone) to 3.94 ± 0.33 µM (n-humulone). Overall, α-acids showed a strong inhibition with selectivity (115–137 fold) for AKR1B10. The results presented herein characterise hop-derived α-acids as a promising basis for the development of novel and selective AKR1B10-inhibitors.


Results and Discussion
In recent research hop-derived prenylflavonoids, including the most prominent hop-compounds xanthohumol and 8-prenylnaringenin, have been subject to a variety of studies in order to elucidate the beneficial effects of these substances in certain disease models. However, experimental data on the biological interaction potential of hop bitter acids, such as (iso-)-α-acids, are relatively scarce [4]. This is especially true for non-isomerized α-acids, which are up to 10-fold (≈10 µM) enriched in many late-or dry-hopped types of beer [28]. Lately, these techniques have become more prominent in the craft beer industry, which, in the future, might lead to increased plasma levels of α-acids following the consumption of certain types of beer.
In this study, mixtures of iso-α-acids and α-acids and the three purified α-acids, compound 1, 2 and 3, were evaluated as inhibitors of the three related human aldo-keto reductases AKR1A1, AKR1B1 and AKR1B10. For comparability reasons and in order to investigate selectivity, DL-glyceraldehyde served as a common test substrate (Table 1-3). Compared to the iso-α-acid mixture, the mixture of α-acids showed superior inhibitory effects with respect to all enzymes tested (Table 1). A slight selectivity for AKR1B1 was observed with the iso-α-acid mixture (Table 1). Table 1. IC50 values of an iso-α-acid solutions and an α-acid mixture for the respective reductases. IC50 values are presented as mean ± SD of at least three experiments.

Results and Discussion
In recent research hop-derived prenylflavonoids, including the most prominent hop-compounds xanthohumol and 8-prenylnaringenin, have been subject to a variety of studies in order to elucidate the beneficial effects of these substances in certain disease models. However, experimental data on the biological interaction potential of hop bitter acids, such as (iso-)-α-acids, are relatively scarce [4]. This is especially true for non-isomerized α-acids, which are up to 10-fold (≈10 µM) enriched in many late-or dry-hopped types of beer [28]. Lately, these techniques have become more prominent in the craft beer industry, which, in the future, might lead to increased plasma levels of α-acids following the consumption of certain types of beer.
In this study, mixtures of iso-α-acids and α-acids and the three purified α-acids, compound 1, 2 and 3, were evaluated as inhibitors of the three related human aldo-keto reductases AKR1A1, AKR1B1 and AKR1B10. For comparability reasons and in order to investigate selectivity, DL-glyceraldehyde served as a common test substrate (Tables 1-3). Compared to the iso-α-acid mixture, the mixture of α-acids showed superior inhibitory effects with respect to all enzymes tested (Table 1). A slight selectivity for AKR1B1 was observed with the iso-α-acid mixture (Table 1). However, AKR1B10 inhibition was up to 115 (ratio AKR1A1/AKR1B10)-137 times (ratio AKR1B1/ AKR1B10) stronger than inhibition of AKR1A1 and AKR1B1, respectively, when a mixture of α-acids was applied (Table 3). Table 3. AKR1B10 selectivity of the α-acid-mixture and its isolated compounds expressed as IC 50 -ratios of AKR1A1/AKR1B10 and AKR1B1/AKR1B10.

Compounds
Ratio AKR1A1/AKR1B10 Ratio AKR1B1/AKR1B10 α-acid mixture 115 137 Shindo et al. [29] report on the inhibitory effect of iso-α-acids on AKR1B1 at lower concentrations (48% inhibition at 33 µg/mL). IC 50 values of iso-α-acids for AKR1B1 in the present study were somewhat higher (100.30 ± 6.03 µg/mL) than reported by Shindo et al. The observed discrepancy to the present study might have been due to the source and quality of the inhibitor as well as due to different purification conditions of the recombinant enzyme. In the present study a 30% (w/w) prediluted standardised solution of iso-α-acids produced from CO 2 hop extract has been used, whereas Shindo et al. used international calibration standard iso-α-acids. Moreover, the concentrations of the single iso-α-acid congeners in the mixture were not further evaluated in both studies and might have influenced the respective IC 50 values as well. Unlike in the present study, Shindo et al. used recombinant AKR1B1 from a eukaryotic expression system, which might have affected the binding behaviour of the inhibitor through posttranslational modifications that are not present in an enzyme derived from the expression system used in this study.
Additionally, potential synergistic effects with other substances found in the hop extract could have also affected AKR1B1 activity. In fact, hop-specific compounds such as xanthohumol, isoxanthohumol and 8-prenylnaringenin have been reported to be strong inhibitors of AKR1B1 and its related reductase AKR1B10 [30].
Due to the promising results of the α-acid extract, AKR-inhibition of the single compounds was of particular interest. UHPLC analysis of the extract showed three dominating peaks ( Figure 2) corresponding to the major α-acids (n-humulone (3), cohumulone (2) and adhumulone (1)) found in hops [28]. Thereupon, α-acids were separated by preparative column chromatography, yielding compound 2 as well as a mixture of compounds 1 and 3 in a first step. Subsequent semi-preparative HPLC led to the isolation of the remaining two main compounds, 1 and 3. All three substances were further analysed for their inhibitory potential.
With respect to the mechanism of inhibition, Zhang et al. suggested a more accessible anionic site at TRP112, allowing the entrance of more bulky and rigid inhibitors at the broader active site of AKR1B10 [33]. Molecular docking experiments have been performed for the three reductases in order to clarify the binding behaviour and specificity of the respective inhibitors. Even though in silico analyses indicated a strong inhibition at the active sites of the enzymes, the mechanism favouring AKR1B10 inhibition over inhibition of the other two enzymes could not ultimately be resolved (data not shown).
With respect to the mechanism of inhibition, Zhang et al. suggested a more accessible anionic site at TRP112, allowing the entrance of more bulky and rigid inhibitors at the broader active site of AKR1B10 [33]. Molecular docking experiments have been performed for the three reductases in order to clarify the binding behaviour and specificity of the respective inhibitors. Even though in silico analyses indicated a strong inhibition at the active sites of the enzymes, the mechanism favouring AKR1B10 inhibition over inhibition of the other two enzymes could not ultimately be resolved (data not shown).
Among the natural-based derivatives serving as selective AKR1B10 inhibitors, hop-derived α-acids have not been investigated so far. For the substances tested in this study, non-competitive and competitive modes of inhibition have been observed (Table 4, Figures 3 and 4). With regard to the inhibition of AKR1B10 mediated farnesal reduction (K M = 5 µM), compound 3 was the best inhibitor with a relatively low K i (3.94 ± 0.33 µM) when compared to its congeners, compound 2 (K i = 16.53 ± 1.74 µM) and 1 (K i = 16.79 ± 1.33 µM).       [34,35]. This was done for all used farnesal and inhibitor concentrations and K was plotted as a function of substrate concentration (lower diagram). This secondary plot shows an increase of K with substrate concentration, which is characteristic for a competitive inhibitor (adhumulone). In the case of a non-competitive inhibitor K would be independent of substrate concentration (humulone and cohumulone). Based on the IC 50 and K i values stated herein, inhibition of AKR1B10 through α-acids during glyceraldehyde or farnesal reduction, might also be compared to the inhibitory efficiency of unsaturated fatty acids on this enzyme (K i -values ranging from 0.24 to 1.1 µM). Unsaturated fatty acids show a competitive inhibition pattern with a specificity towards AKR1B10 [32]. Accordingly, the mechanism of action suggested by Hara et al. [32], which involves the presence of relatively long chain of carbon-carbon double bonds interacting with the enzyme's active site, might in parts also apply for humulone and its three isoprenoid side chains. However, X-ray diffraction experiments would help to further clarify the actual binding mechanism.
Under physiological conditions, the inhibitory effect might be further strengthened with increasing uptake of lipophilic congeners. In this context, a study by Cattoor et al. reported efficient epithelial absorption of α-acids, which might as well point towards a rather high bioavailability [36,37]. Data on the bioavailability on α-acids in animal models are currently lacking; however, the metabolically relevant concentrations (K i , IC 50 ) stated herein, fall within the bioavailable spectrum of iso-α-acids reported by others: a study on the bioavailability of iso-α-acids in rabbits report of cumulative iso-α-acid concentrations between 7 and 20 µM [37]. These concentrations seem sufficient for having an impact on AKR1B10-mediated farnesal reduction. In conjunction with the development of functional foods, there is increasing evidence that prenylation of a target compound raises its bioavailability [38]. As prenylation occurs in both unisomerised and isomerised α-acids, an overall higher bioavailability of these compounds might be expected. In vivo, steadily high concentration levels would be especially important when a competitive mode of inhibition is observed (Table 4, Figures 3 and 4).
In general, research has made great advances in terms of designing new, effective AKR1B10 inhibitors. Unfortunately, though, the clinical safety of their use has in many cases not been evaluated yet [14]. In contrast, hops and hop-derived bitter acids are considered free for consumption and generally recognized as safe for oral intake [28,39,40]. Therefore, α-acids may yield the potential to serve as an alternative basis for the development of AKR1B10-inhibitors.
In conclusion, the results presented in this study identify α-acids as potent inhibitors with a selectivity for AKR1B10 versus homologous AKR1A1 and AKR1B1. Of the three α-acid congeners tested, inhibition by compound 3 showed the strongest inhibition. Moreover, there is evidence of isoprenoid side chains tending to affect the binding behaviour of AKR1B10 inhibitors. With regard to AKR1B10 selectivity, our results provide a structural basis for the development of future QSAR models and new drugs/inhibitors targeting cancers characterized by AKR1B10-specific actions or AKR1B10 upregulation.

Chemicals and Reagents
Organic solvents for chromatography, MS grade water and MS grade formic acid were obtained from VWR (Darmstadt, Germany). Organic solvents used for preparative, semi-preparative and analytical chromatography were of gradient grade quality and water was bi-distilled water. Solvents used for LC-MS analyses were of MS grade quality. Formic acid used for chromatography was of MS grade quality. NADPH was obtained from Carl Roth GmbH & Co. KG (Karlsruhe, Germany). DL-glyceraldehyde and farnesal were purchased from Sigma-Aldrich Co. (St. Louis, MO, USA). A standardised solution of iso-α-acids produced from CO 2 hop extract (30% w/w) was obtained from Barth Haas UK Limited (Tonbridge, UK). Mixtures of α-acids were kindly provided by Dr. Martin Biendl (Hopsteiner-HHV GmbH, Mainburg, Germany).

Preparation of Recombinant Proteins
The carbonyl-reducing enzymes AKR1A1, AKR1B1, AKR1B10 were prepared in an Escherichia coli expression system according to previously published methods: plasmids of AKR1A1 and AKR1B1 were friendly gifts from Prof. Dr. Vladimir Wsol [42] and Dr. Nina Kassner; information about production and purification of AKR1B10 [19] has been published before (sequences of all obtained plasmids containing the specific inserts were verified by sequencing (MWG Eurofins)). The plasmids were then transformed in E. coli BL21 (DE3) cells. For overexpression of 6× His-tagged enzymes, a 400 mL culture (containing the appropriate antibiotic; plasmid dependent) was grown to optical density of 0.6 at 600 nm at 37 • C. Expression was induced by adding isopropyl-1-thio-galactopyranoside to the culture medium (final concentration of 1 mM). After 3 h, cells were harvested by centrifugation (6000× g, 15 min) and resuspended in 20 mL PBS-I buffer (20 mM Na 2 H 2 PO 4 , 500 mM NaCl, 10 mM imidazole, 10% v/v glycerol, 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 respective enzymes were purified using Ni-affinity chromatography (ÄKTA-Purifier; Amersham Pharmacia, Uppsala, Sweden) using PBS-II buffer (20 mM Na 2 H 2 PO 4 , 500 mM NaCl, 500 mM imidazole, 10% v/v glycerol, pH 7.4). 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. For inhibition studies, stock solutions of inhibitors were prepared in H 2 O (iso-α-acid mixture) and DMSO (α-acid mixture and compounds 1-3 purified from the same mixture). The final concentration of DMSO in the assay was ≤ 1% and did not affect enzyme activity. When collecting data for dose-response curves initial velocities of DL-glyceraldehyde or farnesal reduction (substrate concentration at K M ) 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%.

Determination of Inhibition Parameters Using Test Substrates
Initially, the half maximal inhibitory concentrations (IC 50 values) were determined for each inhibitor in presence of each enzyme, using the shared substrate DL-glyceraldehyde (set to their specific K M ; 3.6 mM, 50 µM and 4 mM for AKR1A1, AKR1B1 and AKR1B10, respectively) to assess specificity amongst 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 GraphPad Prism6 (GraphPad Software Inc., La Jolla, CA, USA). Whenever tight-binding inhibition was observed, the inhibition constant Ki was determined by fitting inhibition data to the Morrison equation [43]. In order to verify the inhibitory potency, farnesal as an enzyme-specific physiological substrate for AKR1B10 (farnesal; K M = 5 µM) was used to determine inhibition parameters. Enzyme inhibition parameters were assayed as described above. The inhibition mechanism of each compound for AKR1B10 was analysed by plotting IC 50 -values at different substrate concentrations (at least five inhibitor and substrate concentrations) [43,44]. All data obtained were plotted and analysed using GraphPad Prism6 (GraphPad Software Inc., La Jolla, CA, USA).