Aspartate/asparagine-β-hydroxylase: a high-throughput mass spectrometric assay for discovery of small molecule inhibitors

The human 2-oxoglutarate dependent oxygenase aspartate/asparagine-β-hydroxylase (AspH) catalyses the hydroxylation of Asp/Asn-residues in epidermal growth factor-like domains (EGFDs). AspH is upregulated on the surface of malign cancer cells; increased AspH levels correlate with tumour invasiveness. Due to a lack of efficient assays to monitor the activity of isolated AspH, there are few reports of studies aimed at identifying small-molecule AspH inhibitors. Recently, it was reported that AspH substrates have a non-canonical EGFD disulfide pattern. Here we report that a stable synthetic thioether mimic of AspH substrates can be employed in solid phase extraction mass spectrometry based high-throughput AspH inhibition assays which are of excellent robustness, as indicated by high Z’-factors and good signal-to-noise/background ratios. The AspH inhibition assay was applied to screen approximately 1500 bioactive small-molecules, including natural products and active pharmaceutical ingredients of approved human therapeutics. Potent AspH inhibitors were identified from both compound classes. Our AspH inhibition assay should enable the development of potent and selective small-molecule AspH inhibitors and contribute towards the development of safer inhibitors for other 2OG oxygenases, e.g. screens of the hypoxia-inducible factor prolyl-hydroxylase inhibitors revealed that vadadustat inhibits AspH with moderate potency.

The quality and robustness of the assay was validated by a screening of a compound library of known bioactive small-molecules leading to the identification of novel small-molecule AspH inhibitors. The small-molecule pyridine-2,4-dicarboxylic acid (2,4-PDCA) was identified as a potent AspH inhibitor; a crystal structure of AspH in complex with 2,4-PDCA validates its proposed inhibition mode.
2,4-PDCA was used as a positive control in all subsequent dose-response experiments; low standard deviations (SD) throughout all measurements with 2,4-PDCA were observed (n = 10). The binding mode of 2,4-PDCA to AspH was then investigated by crystallography, in part to provide evidence that the SPE-MS AspH inhibition assay does not afford false-positive inhibition results. crystallographic validation of the AspH inhibition assay. A crystal structure of His 6 -AspH 315-758 bound to Mn, 2,4-PDCA, and a synthetic AspH substrate peptide (hFX-EGFD1 86-124 -4Ser, Supporting Figure S1) mimicking the full-length EGFD1 of hFX was obtained (AspH:2,4-PDCA, Fig. 3a). His 6 -AspH 315-758 crystallized in the P2 1 2 1 2 1 space group (2.24 Å resolution), identical to that obtained when using NOG instead of 2,4-PDCA as an inhibitor (AspH:NOG; PDB ID: 5JQY) 34 . The AspH oxygenase domain is comprised of a distorted double-stranded β-helix fold bearing two Fe(II) binding histidine residues; 33, 34 The AspH TPR domain appears to be essential for productive substrate binding as it provides a hydrophobic pocket for conserved EGFD tyrosine or phenylalanine residues in the substrate consensus sequence to bind (Fig. 3a) 34 .
Analysis of the electron density confirms that 2,4-PDCA replaces 2OG and mimics 2OG binding in the AspH active site; it coordinates to the Mn ion, substituting for the Fe(II) cofactor due to crystallisation purposes, in a bidentate fashion by its pyridyl-nitrogen and one of its C-2 carboxylate oxygen atoms ( Fig. 3c and d), consistent with other crystal structures of 2OG oxygenases in complex with 2,4-PDCA (Supporting Figures S2 and S3). 2,4-PDCA is positioned to interact via its C-4 carboxylate oxygens with the 2OG C-5 carboxylate binding residues, Arg735 and Ser668 (at least one Lys-or Arg-residue and an alcohol bearing residue are typically involved in 2OG C-5 50 and 2,4-PDCA C-4 carboxylate binding, Supporting Figure S3). 2,4-PDCA is also positioned to interact with the AspH active site, through its non-metal coordinating C-2 carboxylate oxygen with the N ε -imidazole nitrogen of His690 (2.8 Å) and its metal coordinating oxygen with a guanidyl nitrogen of Arg688 (2.8 Å); the presence of a non-metal ion chelating histidine at the active site of 2OG oxygenases is not conserved, but is precedented 51 . The C-1 carboxylate of NOG (and likely 2OG) can make an analogous interaction with His690, but the rigid cyclic nature of 2,4-PDCA may favour it on entropic grounds. Overall, analysis of the crystal structures reveals that 2,4-PDCA and NOG bind similarly to AspH by replacing 2OG in the active site rationalizing their activities against AspH and validating the SPE-MS AspH inhibition assay.
Validation of the robustness of the AspH inhibition assay. Having confirmed that the SPE-MS AspH assay can be applied to identify small-molecule AspH inhibitors, we validated it by profiling AspH against the library of pharmacologically active compounds (LOPAC, Sigma-Aldrich) extended by selected Pfizer-developed bioactive compounds (in total 1370 small-molecules). Initially, the compounds were dispensed on five 384-well plates at a fixed concentration (20 μM), and their inhibition of AspH analysed using the SPE-MS high-throughput assay. The good data quality is reflected by high signal-to-noise (S/N) and signal-to-background (S/B) ratios (Fig. 4a). The five Z'-factors were >0.6 for each plate, indicating, by definition, a highly robust and stable assay of excellent quality (Fig. 4b) 52 .
At a fixed compound concentration (20 μM), 48 out of the 1370 bioactive compounds inhibited >95% AspH activity (see Supporting Table S1 and Supporting Data Sheet). Several of the most potent inhibitors are either flavonoids or contain (metal ion chelating) catechol-moieties and have been identified to inhibit 2OG oxygenases in previous screens 53,54 , including the LOPAC screen 55 . These compounds likely inhibit 2OG oxygenases by reducing free Fe(II) concentration in solution (through chelation), by changing the redox equilibrium in vitro, and/or by directly binding the enzymes [53][54][55] . Out of the initial 48 hits, 14 representative compounds covering the different compound classes were selected to independently determine their dose-responses (IC 50 -values) against AspH ( Table 1). All AspH inhibition assays performed to generate dose-response curves were of high quality (Z'-factors >0.5, Fig. 4c).
Many flavonoids are part of daily human nutrition and some are taken as health supplements; their impact on modulating human biology is thus of general public health interest. The flavonoid natural product quercetin, which is an ubiquitous part of dietary intake 56,57 and a reported 2OG oxygenase inhibitor 54 , inhibits AspH activity efficiently (IC 50 ~ 0.25 μM; Table 1, entry 10). Considering that quercetin uptake has been correlated to inhibit cancer progression [58][59][60] , it is possible that the inhibition of AspH accounts for some of its biological effect, however, it should be noted that the effects of flavonoids are likely pleiotropic.
The natural product caffeic acid, which is also present in human nutrition and displays a variety of biological effects when administered to humans 61 , inhibits AspH with a similar potency to quercetin (Table 1, entry 11). Caffeic acid is a reported 2OG oxygenase inhibitor; 53,54 its mechanism of action could either involve displacement of 2OG in the active site, covalent binding to AspH through reaction of its Michael acceptor with nucleophilic protein residues, Fe(II) chelation by its catechol moiety, or modulation of the redox environment by radical scavenging 61 .
Apart from natural products, synthetic bioactive molecules were identified to inhibit AspH catalysis. The potent c-RAF1 kinase inhibitor GW5074 62 , the ionophore calcimycin (A23187) 63 , the inhibitor SCH-202676 effecting agonist and antagonist binding to G-protein-coupled receptors 64 , the leukotriene biosynthesis inhibitor MK-886 65 , and the ion channel antagonist capsazepine 66 all inhibited AspH activity with SCH-202676 exhibiting the highest potency (IC 50 ~ 0.5 μM; Table 1, entry 4). The κ-opioid receptor antagonist GNTI 67 was recognized as potent AspH inhibitor in the initial LOPAC screen (100% inhibition; Table 1, entry 1), but was inactive as a freshly prepared DMSO stock solution, suggesting that a decomposition product may cause the originally observed inhibition.
Several active pharmaceutical ingredients (APIs) approved for human therapy were AspH inhibitors. For instance, the tetracycline-derived antibiotic tigecycline 68 and sildenafil, the API in Viagra 69 , inhibit AspH with micromolar potency (Table 1, entries 13 and 9). The antibiotic cephalosporin C zinc salt 70 is a potent AspH inhibitor (IC 50 ~ 0.07 μM; Table 1, entry 8). The latter observation is likely due to Zn(II) substituting for Fe(II) at the AspH active site, as independently determined dose-response of Zn(II) ions suggest (see below) 38 ; other metal-free cephalosporin antibiotics did not inhibit AspH. The anticancer drug PAC-1 activates procaspase-3 by sequestering Zn(II) and presumably inhibits AspH by a similar sequestration of Fe(II) 71 . Candesartan cilexetil, an angiotensin II type 1 receptor antagonist 72 , inhibited AspH activity only in form of its prodrug cilexetil ester, a modification used to increase the bioavailability of the bioactive carboxylic acid drug; no inhibition of the bioactive candesartan containing a free carboxylic acid instead of the cilexetil ester was observed (Table 1, entry 5). This observation likely reflects involvement of the cilexetil ester, which might chelate Fe(II); binding to AspH by candesartan cilexetil was not observed by analyzing the shift of AspH melting temperature (T m ) in the presence of the small-molecule using differential scanning fluorimetry (ΔT m <0.3 °C, Supporting Figure S4). The structure of benserazide (Ro 4-4602, Table 1, entry 14) 73 , which is administered in combination with levodopa, is similar to catechol-containing compounds and might thus inhibit AspH by chelating Fe(II) 53,54 .
In summary, the LOPAC screen results highlight the quality and robustness of the SPE-MS AspH inhibition assay and led to the identification of both natural product as well as synthetic potent and novel AspH inhibitors. (2020) 10:8650 | https://doi.org/10.1038/s41598-020-65123-9 www.nature.com/scientificreports www.nature.com/scientificreports/ Our assay was then applied to investigate the effects of selected reported 2OG oxygenase inhibitors on AspH activity.

Inhibition of AspH by reported 2OG oxygenase inhibitors.
Clinical interest in modulating the activity of 2OG oxygenases has spurred the development of small-molecule 2OG oxygenase inhibitors, most of which are active site Fe(II) chelators and 2OG competitors; some of which are now APIs in approved human therapeutics 28,29 . A selection of reported 2OG oxygenase inhibitors including various metal salts were profiled against AspH to identify potential lead structures for the development of potent and selective AspH inhibitors and inform on the selectivity, and thus potential side effects, of 2OG oxygenase inhibitors in clinical trials and clinical use.
Initially, the impact of different metal ions on AspH activity was examined, in part because the unusual geometry of the AspH active site suggests that Fe(II) binding to AspH might be different to other 2OG oxygenases as Fe(II) is only bound by two histidine residues (His679, His725; Fig. 3) rather than by the typical three protein ligands (HXD/E…H) 33, 34 . Co(II) ions have also been used for treatment of anaemia 74 , in manner that might be mediated via hypoxia-inducible transcription factor (HIF) prolyl hydroxylase inhibition 75,76 . Indeed, Zn(II), Ni(II), and Co(II) inhibited AspH activity, with Zn(II) exhibiting the most pronounced effect (IC 50 ~ 0.05 μM, Table 2) consistent with observed AspH inhibition by cephalosporin C zinc salt ( Table 1, entry 8). Ni(II) and Co(II) are less potent than Zn(II), but still inhibit AspH at sub-micromolar concentrations ( Table 2). Inhibition of AspH is reduced approximately 100 fold for Mn(II) compared to Zn(II) ( Table 2). Ca(II) ions do not show   Table 1. Inhibition of AspH by selected small-molecules in the LOPAC screening set. a) The complete screening results are shown in the Supporting Data Sheet; b) mean average of two independent runs (n = 2; mean ± SD). AspH inhibition assays were performed as described in the Methods Section using 50 nM His 6 -AspH 315-758 and 1.0 μM hFX-CP 101-119 (Fig. 1d) as a substrate; c) candesartan was used in the bioactive acid form rather than as its prodrug cilexetil ester form. www.nature.com/scientificreports www.nature.com/scientificreports/ any effect on AspH activity, at least with the tested thioether substrate hFX-CP 101-119 ( Table 2). The initial results suggest that AspH might be much more sensitive towards the presence of metal ions compared to many other 2OG oxygenases (Table 2), e.g. HIF-α prolyl hydroxylases 1-3 (PHD1-3) 77 , factor inhibiting HIF (FIH) 78 , taurine/αKG dioxygenase (TauD) 79 , and the JmjC lysine-specific demethylases KDM3A 80,81 , KDM4A/JMJD2A 77 , and KDM4E/JMJD2E 77 , though the differences could also relate to the different Fe(II) concentrations and assay techniques used.

Discussion
The observation that increased levels of AspH on the surface of cancer cells correlates with enhanced tumour invasiveness 3,4,14 coupled with the observations that AspH is upregulated in many tumours 8,[14][15][16][17] and is hypoxically regulated 41,42 , renders AspH an interesting potential cancer target. The exact mechanism of how AspH promotes tumour invasiveness is poorly understood, though there is some evidence it is mediated by the notch receptor and its ligands, which contain multiple EGFDs bearing the consensus sequence for AspH-catalysed hydroxylation 8,9,23,32 . The identification of potent and selective small-molecule AspH inhibitors should help dissect the biological roles of AspH and investigate it as a drug target. However, work towards this has been limited by lack of an efficient AspH assay.
Employing the semi-automated high-throughput RapidFire sampling robot, the library of pharmacologically active compounds (LOPAC) was screened, as was done for another 2OG oxygenase, KDM4E (JMJD2E), employing a fluorescence based assay 55 . The stability and robustness of the AspH assay was highlighted by excellent Z'-factors (Fig. 4); the assay only lacked accuracy when strongly ionizing small-molecules suppressed the ionization of the hFX-CP 101-119 substrate. Both natural products and synthetic bioactive molecules, some of which are APIs of approved human therapeutics, were identified from the LOPAC set as potent AspH inhibitors (Table 1,  Supporting Table S1, and Supporting Data Sheet). In general, AspH and KDM4E were inhibited by structurally similar LOPAC compounds, including reported redox-active or metal ion chelators. More compounds were identified that inhibit AspH than KDM4E, possibly reflecting the different assay conditions used (e.g. use of 2 μM Fe(II) for AspH; 10 μM Fe(II) for KDM4E). The potential sensitivity of AspH towards redox active compounds might in part reflect its nature as an ER protein bearing one disulfide and four free cysteine residues in its oxygenase domain 34 . It should be noted that the results of the SPE-MS AspH inhibition assay alone do not define the mechanism of action of the identified AspH inhibitors. Many small-molecules from the obtained LOPAC hit-list likely inhibit AspH by modulating the redox equilibrium of the reaction or by reducing the concentration of available Fe(II). Such compounds can be identified by using a combination of SPE-MS and biophysical techniques such as crystallography (Fig. 3 and Supporting Figures S2 and S3), DSF (Supporting Figure S4), non-denaturing MS, NMR or surface plasmon resonance (SPR)/bio-layer interferometry (BLI) as counterscreens.
The AspH active site geometry is different than that of other human 2OG dependent hydroxylases as the Fe(II) cofactor is bound by only two ligands (His679, His725; Fig. 3) rather than the more typical triad of ligands (HXD/E…H) 33,34 . However, under our assay conditions, the experimentally determined K m value of AspH for Fe(II) is in the range of other 2OG oxygenases 36 . Of the tested metal ions, Zn(II) ions inhibited AspH activity with the highest potency (IC 50 ~ 0.05 μM, Table 2), in agreement with prior studies using native bovine AspH 38 . The inhibitory effect of Zn(II), Ni(II), Co(II), and Mn(II) on AspH activity seems to be enhanced compared to other 2OG oxygenases (Table 2) including the asparaginyl-and aspartyl-residue hydroxylase FIH 78 . Care should be taken in drawing firm conclusions on metal ion binding based on turnover assays, due to differences in conditions employed, e.g. a relatively low Fe(II) concentration was used in our AspH assay and our AspH construct bears a His 6 -tag. Nonetheless, the results suggest relative inhibition of 2OG oxygenases by metal ions may vary in cells to Scientific RepoRtS | (2020) 10:8650 | https://doi.org/10.1038/s41598-020-65123-9 www.nature.com/scientificreports www.nature.com/scientificreports/ a different degree. This is of biomedical relevance as Co(II) ions have been used to treat anaemia, potentially via inhibition of the HIF prolyl hydroxylases [74][75][76] .
The presence of Ca(II) ions does not perturb AspH activity under our assay conditions, an observation of potential significance as full-length AspH contains at least one Ca(II)-binding EF hand domain 34 , most AspH substrate EGFDs contain Ca(II)-binding sites around the Asp/Asn-hydroxylation site, and AspH isoforms lacking the catalytic oxygenase domain are involved in cellular Ca(II) homeostasis. However, it should be noted, by contrast with the canonical EGFD domain fold, that the non-canonical disulfide pattern thioether substrate analogue used by us, does not bind Ca(II) ions. Given that the canonical and non-canonical EGFD disulfide forms likely interconvert in vivo, the regulation of AspH by Ca(II) ions is a possibility.
Inhibiting human 2OG oxygenases linked to disease by small-molecules is a successful strategy for the development of novel therapeutics 29,33 , as shown by work on inhibition of the HIF prolyl hydroxylases 108 . Some of the JmjC hydroxylases and KDMs are also being investigated as medicinal chemistry targets. Given the roles of many 2OG oxygenases, including AspH, in important biological processes, achieving inhibitor selectivity is likely important, especially when treating long term diseases such as anaemia. We thus tested several APIs of approved and investigated therapeutics against AspH ( Table 3). APIs of approved therapeutics such as roxadustat 102 and meldonium 101 which presumably inhibit their target 2OG oxygenase by binding to its active site, are weak inhibitors of AspH, likely reflecting structural differences distinguish the binding pockets of their 2OG oxygenase targets (the PHDs for roxadustat and BBOX for meldonium). Among the other APIs currently in clinical trials for treating anemia in patients suffering from chronic kidney disease, only the HIF prolyl hydroxylase inhibitor vadadustat inhibited AspH activity substantially ( Table 3, entry 19). The therapeutic APIs deferoxamine, deferasirox, deferiprone, and ciclopirox which exert their biologic function indirectly through the chelation of Fe(II), also inhibit AspH substantially. The results presented here suggest that AspH inhibition should be a consideration in analysing the side effects of the aforementioned APIs.
Overall, the current study should enable the development of selective small-molecule AspH inhibitors which will be useful as chemical probes that will complement biological approaches for inhibiting AspH, as well as helping develop inhibitors selective for other 2OG oxygenases not inhibiting AspH. The natural products and synthetic small-molecules AspH inhibitors identified may be useful as lead structures for initiating AspH inhibitor development programs. Recombinant AspH production and purification. N-Terminally His 6 -tagged human AspH 315-758 (His 6 -AspH 315-758 ) was produced in E. coli BL21 (DE3) cells using a pET-28a(+) vector as previously reported 34,36 . After cell lysis, AspH was purified by Ni(II)-affinity chromatography (HisTrap HP column, GE Healthcare; 1 mL/ min flow rate) and size-exclusion chromatography (HiLoad 26/60 Superdex 75 pg 300 mL column; 1 mL/min) using an ÄKTA pure machine (GE Healthcare) as reported. AspH was >95% pure by SDS-PAGE and MS analysis and had the anticipated mass as reported 34 , it was stored in 50 mM HEPES buffer (pH 7.5, 150 mM NaCl) at a concentration of 125 μM at −78 °C; fresh aliquots were used for all biochemical experiments.
Solutions of the bioactive small-molecules (100% DMSO) were dry dispensed across 384-well polypropylene assay plates (Greiner) using an ECHO 550 acoustic dispenser (Labcyte). DMSO and 2,4-PDCA were used as negative and positive controls. The DMSO concentration was kept constant at 0.5% v/v throughout all experiments (using the DMSO backfill option of the acoustic dispenser). The initial screening of the LOPAC was performed at a fixed compound concentration (20 μM). For dose-response experiments, the AspH inhibitors were dry dispensed in an approximately three-fold and 11-point dilution series using the acoustic dispenser (100 μM or 10 μM top concentration). 2,4-PDCA was used as a positive control; its IC 50 -value was determined on each assay plate to confirm the assay quality. Each reaction was performed in technical duplicates in adjacent wells on the assay plates; additionally, assays were performed in two independent duplicates on different days using different DMSO inhibitor solutions.
MS-analyses were performed using a RapidFire RF 365 high-throughput sampling robot (Agilent) attached to an iFunnel Agilent 6550 accurate mass quadrupole time-of-flight (Q-TOF) mass spectrometer operated in the positive ionization mode. Assay samples were aspirated under vacuum for 0.4 s and loaded onto a C4 solid phase extraction (SPE) cartridge. After loading, the C4 SPE cartridge was washed with 0.1% v/v aqueous formic acid to remove non-volatile buffer salts (5 s, 1.5 mL/min). The peptide was eluted from the SPE cartridge with 0.1% v/v aqueous formic acid in 85/15 v/v acetonitrile/water into the mass spectrometer (5 s, 1.25 mL/min) and the SPE cartridge re-equilibrated with 0.1% v/v aqueous formic acid (1 s, 1.25 mL/min). The mass spectrometer parameters were: capillary voltage (4000 V), nozzle voltage (1000 V), fragmentor voltage (365 V), gas temperature (280 °C), gas flow (13 L/min), sheath gas temperature (350 °C), sheath gas flow (12 L/min). The m/z + 2 charge states of the cyclic peptide (substrate) and the hydroxylated cyclic peptide (product) were used to extract ion chromatogram data, peak areas were integrated using RapidFire Integrator software (Agilent). The data were exported into Microsoft Excel and used to calculate the % conversion of the hydroxylation reaction using the equation: % conversion = 100 ×(integral product cyclic peptide) / (integral substrate cyclic peptide + integral product cyclic peptide). Normalized dose-response curves (2,4-PDCA and DMSO controls) were obtained from the raw data by non-linear regression (GraphPad Prism 5) and used to determine IC 50 -values. The standard deviation (SD) of two independent IC 50 determinations (n = 2) was calculated using GraphPad Prism 5. Signal-to-noise (S/N) and signal-to-background (S/B) ratios 52 as well as Z'-factors 52 and minimum significant ratios (MSR) 109 were calculated according to the cited literature using Microsoft Excel.
Structure solution and refinement. The AspH crystal structure was determined by molecular replacement (MR) using the AutoMR (PHASER 111 ) subroutine in PHENIX 112 . The search model used for molecular replacement was based on PDB-ID 5JQY 34 ; the structural model was improved by iterative cycles of manual re-building in COOT 113 and crystallographic refinement in PHENIX (Supporting Table S3).

Data availability
The data for the AspH:2,4-PDCA crystal structure are deposited in the protein databank with the PDB accession code 5JTC.