Small molecule-protein interactions exemplified on short-chain dehydrogenases/reductases

The short-chain dehydrogenase/reductase (SDRs) family represents one of the largest enzyme superfamilies, with over 80 members in the human genome. Even though the human genome project has sequenced and mapped the entire human genome, the physiological functions of more than 70% of all SDRs are currently unexplored or insufficiently characterized. To start to fill this gap, the present thesis aimed to employ a combination of molecular modeling approaches and biological assessments for the identification and characterization of novel inhibitors and/or potential substrates of different SDRs.
Due to their involvement in steroid biosynthesis and metabolism, SDRs are potential targets of endocrine disrupting chemicals (EDCs). To test the use of pharmacophore-based virtual screening (VS) applications and subsequent in vitro evaluation of virtual hits for the identification and characterization of potential inhibitors, 11β-hydroxysteroid dehydrogenase 2 (11β-HSD2) was selected as an example. 11β-HSD2 has an important role in the placenta by inactivating cortisol and protecting the fetus from high maternal glucocorticoid levels. An impaired placental 11β-HSD2 function has been associated with altered fetal growth and angiogenesis as well as a higher risk for cardio-metabolic diseases in later life. Despite this vital function, 11β-HSD2 is not covered in common off-target screening approaches. Several azole fungicides were identified as 11β-HSD inhibitors amongst approved drugs by testing selected virtually retrieved hits for inhibition of cortisol to cortisone conversion in cell lysates expressing recombinant human 11β-HSD2. Moreover, a significant structure-activity relationship between azole scaffold size, 11β-HSD enzyme selectivity and potency was observed. The most potent 11β-HSD2 inhibition was obtained for itraconazole (IC50 139 ± 14 nM), for its active metabolite hydroxyitraconazole (IC50 223 ± 31 nM), and for posaconazole (IC50 460 ± 98 nM). Interestingly, substantially lower inhibitory 11β-HSD2 activity of these compounds was detected using mouse and rat kidney homogenate preparations, indicating species-specific differences. Impaired placental 11β-HSD2 function exerted by these compounds might, in addition to the known inhibition of P-glycoprotein efflux transport and cytochrome P450 enzymes, lead to locally elevated cortisol levels and thereby could affect fetal programming.
Successful employment of pharmacophore-based VS applications requires suitable and reliable in vitro validation strategies. Therefore, the following study addressed the re-evaluation of a potential EDC, the widely used flame retardant tetrabromobisphenol A (TBBPA), on glucocorticoid receptor (GR) and androgen receptor (AR) function. TBBPA was reported earlier in yeast-based reporter assays to potently interfere with GR and moderately with AR function. Human HEK-293 cell-based reporter assays and cell-free receptor binding assays did not show any activity of TBBPA on GR function, which was supported by molecular docking calculations. The antiandrogenic effect, however, could be confirmed, although less pronounced than in the HEK-293 cell system. Nevertheless, the evaluation of the relevant concentrations of an EDC found in the human body is crucial for an appropriate safety assessment. Considering the rapid metabolism of TBBPA and the low concentrations observed in the human body, it is questionable whether relevant concentrations can be reached to cause harmful effects. Thus, it is vital to take the limitations of each testing system including the distinct sensitivities and specificities into account to avoid false positive or false negative results.
To extend the applications of in silico tools with demonstrated proof-of-concept, they were further employed to investigate novel substrate specificities for three different SDR members: the two multi-functional enzymes, 11β-HSD1 and carbonyl reductase (CBR) 1 as well as the orphan enzyme DHRS7.
A role for 11β-HSD1 in oxysterol metabolism by metabolizing 7-ketocholesterol (7kC) has already been described. However, in contrast to the known receptors for 7α,25-dihydroxycholesterol (7α25OHC), i.e. Epstein-Barr virus-induced gene 2 (EBI2), or 7β,27-dihydroxycholesterol (7β27OHC), i.e. retinoic acid related orphan receptor (ROR)γ, no endogenous receptor has been identified so far for 7kC or its metabolite 7β-hydroxycholesterol. To explore the underlying biosynthetic pathways of such dihydroxylated oxysterols, the role of 11β-HSD1 in the generation of dihydroxylated oxysterols was investigated. For the first time, the stereospecific and seemingly irreversible oxoreduction of 7-keto,25-hydroxycholesterol (7k25OHC) and 7-keto,27-hydroxycholesterol (7k27OHC) to their corresponding 7β-hydroxylated metabolites 7β25OHC and 7β27OHC by recombinant human 11β-HSD1 could be demonstrated in vitro in intact HEK-293 cells. Furthermore, 7k25OHC and 7k27OHC were found to be potently inhibited the 11β-HSD1-dependent oxoreduction of cortisone to cortisol. Molecular modeling experiments confirmed these results and suggested competition of 7k25OHC and 7k27OHC with cortisone in the enzyme binding pocket.
For a more detailed enzyme characterization, 11β-HSD1 pharmacophore models were generated and employed for VS of the human metabolome database and the lipidmaps structure database. The VS yielded several hundred virtual hits, including the successful filtering of known substrates such as endogenous 11-ketoglucocorticoids, synthetic glucocorticoids, 7kC, and several bile acids known to inhibit the enzyme. Further hits comprised several eicosanoids including prostaglandins, leukotrienes, cyclopentenone isoprostanes, levuglandins or hydroxyeicosatetraenoic acids (HETEs) and compounds of the kynurenine pathway. The important role of these compounds as well as 11β-HSD1 in inflammation emphasizes a potential association. However, further biological validation is of utmost necessity to explore a potential link.
The closest relative of 11β-HSD1 is the orphan enzyme DHRS7, which has been suggested to act as tumor suppressor. Among others, cortisone and 5α-dihydrotestosterone have been identified as substrates of DHRS7, although effects in functional assays could only be observed at high concentrations that may not be of physiological relevance. Hence, the existence of other yet unexplored substrates of DHRS7 can be assumed, and the generation of homology models to study the structural features of the substrate binding site of DHRS7 was employed. The predictivity of the constructed models is currently limited, due to a highly variable region comprising a part of the ligand binding site but particularly the entry of the binding pocket, and requires further optimizations. Nevertheless, the models generally displayed a cone-shaped binding site with a rather hydrophobic core. This may suggest larger metabolites to be converted by DHRS7. Moreover, the flexible loops surrounding the binding pocket may lead to the induction of an induced fit upon ligand binding. However, further studies are crucial to confirm these findings.
CBR1 is well-known for its role in phase I metabolism of a variety of carbonyl containing xenobiotic compounds. Several endogenous substrates of CBR1 have been reported such as prostaglandins, S-nitrosoglutathione or lipid aldehydes. The physiological relevance of these endogenous substrates, however, is not fully understood. Thus, the physiological roles of CBR1 was further explored by identifying a novel function for CBR1 in the metabolism glucocorticoids. CBR1 was found to catalyze the conversion of cortisol into 20β-dihydrocortisol (20β-DHF), which was in turn detected as the major route of cortisol metabolism in horses and elevated in adipose tissue derived from obese horses, humans and mice. Additionally, 20β-DHF was demonstrated as weak endogenous agonist of the GR, suggesting a novel pathway to modulate GR activation by CBR1-depenent protection against excessive GR activation in obesity.
In conclusion, this thesis emphasized the employment of molecular modeling approaches as an initial filter to identify toxicological relevant compound classes for the identification of potential EDCs and, moreover, as valuable tools to identify novel substrates of multifunctional SDRs and to unravel novel functions for the large majority of yet unexplored orphan SDR members, while carefully considering the limitations of this strategy.