Pyrazolyl-pyrimidones inhibit the function of human solute carrier protein SLC11A2 (hDMT1) by metal chelation

Activity optimization of a pyrazolyl-pyrimidone against the iron transporter SLC11A2 (hDMT1) and mechanistic studies revealed that this class of inhibitors act by metal chelation.


Introduction
Solute carrier proteins (SLCs) control fluxes of ions and other molecules across biological membranes and represent an emerging class of drug targets. [1][2][3] Here, we investigate inhibitors of the human H + -coupled transporter of ferrous iron (Fe 2+ ), SLC11A2 (hDMT1). This transporter is expressed in the intestinal brush border membrane, where it acts as the key mediator of dietary iron uptake. hDMT1 is also linked to pathologies such as hereditary hemochromatosis, β-thalassemia, Parkinson's disease and Alzheimer's disease, highlighting that its pharmacological inhibition may be beneficial to treat human diseases. [4][5][6][7][8][9] While metal chelators have been classically used to treat metal intoxication 10 and neurodegenerative diseases, 11,12 a recently reported highly potent and specific inhibitor of ferroportin (SLC40), a different iron transporter acting on the same pathway as hDMT1, has been shown to have clinical efficacy against β-thalassemia. 13 Two families of small molecule hDMT1 inhibitors have been reported in the literature as the results of high-throughput screening campaigns, namely bis-cationic isothioureas such as dibenzofurans 1 and mesitylene 2, 14 as well as pyrazolyl-pyridine 3 (Fig. 1). 15 In our own investigations on hDMT1, we used the inhibitors mentioned above as seeds for a ligand-based virtual screening campaign guided by 3D-shape and pharmacophore similarity 16 and discovered bis-isothiourea 4 and pyrazolylpyrimidone 5 as two additional hDMT1 inhibitors. 17 Kinetic studies, an X-ray structure of a related brominated bisisothiourea inhibitor in complex with a bacterial analog of the transporter and mutational studies recently showed that bisisothiourea-based compounds act as competitive inhibitors of hDMT1. 18 On the other hand, pyrazolyl-pyrimidone 5, whose small size and better drug-like properties made it an attractive candidate compound, acted as a non-competitive inhibitor. Here we set out to investigate the pyrazolyl series (3 and 5) closer and understand its mechanism of action.

Optimization of pyrazolyl-pyrimidone 5
We prepared twenty new analogs (6-25) of pyrazolyl-pyrimidone 5 by condensing aminoguanidines with 1,3-diketones and ketoesters to form pyrazoles 6-20 and pyrazolones 21-25 using known chemistry (Scheme 1). 14,15,19,20 For pyrazoles the synthesis gave only one product from symmetrical diketones (5-9, 16, 18-20). With non-symmetrical diketones synthesis also mostly yielded a single product. X-ray crystal structures of 13 and 17 showed that the product from non-symmetrical diketones was the isomer with the larger pyrazole substituent pointing away from the aminopyridone. By analogy, the same arrangement was assigned to the other pyrazoles. The synthesis of pyrazolones 21-25 only gave a single product whose structures were confirmed by an X-ray crystal structure of 24 as hydrochloride salt (Fig. 2).
We measured hDMT1 activity using a radioactive iron uptake assay in HEK293 cells stably overexpressing hDMT1 as described previously. 17 Dibenzofuran inhibitor 1 showed only micromolar potency in this assay, which is 5-fold weaker than the originally reported value based on a calcein assay, while the mesitylene inhibitor 2 showed a comparable value to the originally reported inhibition (Table 1). For the present study, we selected dibenzofuran inhibitor 1 as positive control. However, we were unable to determine an IC 50 value for the known pyrazolyl-pyridine 3 due to a lack of inhibition at high compound concentration.
Activity screening of the twenty synthesized analogs of 5 for inhibition of hDMT1 revealed that seven pyrazoles (6,7,13,(16)(17)(18)(19) showed similar or improved inhibition compared to 5. On the other hand, pyrazoles 8-12, 14-15 and 20 as well as pyrazolones 21-25 did not show significant inhibition of iron uptake in our assay. Across all compounds tested, pyrazole 13 stood out as one of the most potent compounds with IC 50 = 1.1 ± 0.01 μM, representing a 10-fold improvement over our initial inhibitor 5 while preserving the original ligand efficiency. Note that the bis-cyclohexane analog 18 was slightly more potent, however at the cost of higher hydrophobicity.
On the other hand, the uptake of radioactive iron in Xenopus oocytes overexpressing hDMT1 was totally inhibited by the bis-thiourea inhibitor 1, while pyrazolyl-pyrimidone 13 only inhibited 50% of the uptake (Fig.  3a). Electrophysiological recordings in this system furthermore showed that only inhibitor 1 blocked hDMT1-induced current with a potency comparable to the iron uptake experiment (IC 50 = 0.14 μM), but that pyrazolyl-pyrimidone 13 had no measurable effect on the recorded currents ( Fig. 3b/c).
Taken together, these experiments showed that 13 was indeed a more potent version of the non-competitive inhibitor 5. However, in contrast to the well-behaved bisisothiourea 1, the inhibitory effect of 13 was only detected in iron uptake experiments and was not observed in electrophysiological recordings. A similar effect had been reported for the literature inhibitor 3. 15

Chemical evidence for metal complex formation
At this stage we suspected that pyrazolyl-pyrimidone 13 might inhibit hDMT1 indirectly by chelating iron, although this  possibility was judged unlikely for the related inhibitor 3 by lack of conclusive evidence. 15 These inhibitors contain an N:NĲ:C)-C:N substructure which is potentially capable of forming a 5-membered ring chelate with metal ions via the two terminal sp 2 nitrogen atoms, although this substructure is present in all the synthesized analogs including those that did not show any inhibition. The formation of such 5-membered ring chelate has been recently established by X-ray crystallography of a ruthenium complex containing both a pyrazolyl-pyridine and a pyrazolyl-pyridone ligand and investigated as C-H oxidation catalyst. 22 Unfortunately, UVvis titration of 13 could not be used to provide any evidence for metal complexation since the UV-vis spectrum did not significantly change upon titration with Fe 2+ and Cd 2+ . Attempts to crystallize the iron complex of 13 as well as investigation by 1 H-NMR and mass spectrometry similarly did not yield any indication of complex formation.
A first indication that pyrazolyl-pyrimidone 13 might interact with divalent metal ions was obtained by isothermal titration calorimetry (ITC) with Cd 2+ , which we used as a redox stable analog of Fe 2+ less prone to hydroxide formation at pH = 7.4 and which is also transported by DMT1. At a pH of 7.4, the ITC data indicated an apparent dissociation constant of K D = 6.4 μM with n = 0.54, consistent with the formation of a 2:1 ligand-metal complex (Fig. 4a). At pH 5.5 under which the hDMT1 transport experiment is carried out,  the ITC data could not be interpreted with certainty. Upon titration of Cd 2+ into a solution of 13, the exothermic signal seems to saturate after the first injections, but later results in a noisy signal, indicating precipitation (Fig. 4b). The less potent pyrazolyl-pyrimidone DMT1 inhibitor 5 also interacted with Cd 2+ at pH 7.4, although binding was much weaker (K D ∼ 250 μM, n = 0.74). A calcein competition assay to detect Fe 2+ chelation at pH 5.5 also indicated a significant level of iron chelation by 13, although the effect was weaker than with the positive control bipyridine or the literature inhibitor 3, and pyrazolylpyrimidone 5 did not show any iron chelation in this assay (Fig. 4c). Tracking the radioactivity of 55 Fe 2+ in solutions after incubation with the various inhibitors indicated that pyrazolyl-pyrimidone 13 as well as the literature inhibitor 3 induced significant precipitation of Fe 2+ at pH 7.4, while bisisothiourea 1, bipyridine or pyrazolyl-pyrimidone 5 had no effect. On the other hand, at pH 5.5, we only observed precipitation with the literature inhibitor 3 corresponding to the pH of the assay with hDMT1 ( Fig. 4d and e). Such precipitation might indicate chelation of iron to form an insoluble complex. Taken together, these data pointed to potential iron chelating abilities of our inhibitors, however the evidence was inconclusive and did not match well with the observed inhibitory potencies.

Cross-inhibition of human SLC11A2 (hDMT1) and SLC39A8 (hZIP8)
Complex formation with Fe 2+ should result in non-selective inhibition of different metal transporters. To test this hypothesis, we investigated whether human SLC39A8 (hZIP8), a transporter of Zn 2+ and Fe 2+ unrelated to SLC11 family, might also be inhibited by our hDMT1 inhibitors. Both SLCs can be tested by inhibition of radioactive iron uptake into . Kinetic parameters were obtained by fitting experimental results to a 4-parameter sigmoidal curve (black line). Compound 13 did not fit to the equation while for compound 1 the obtained IC 50 was 0.14 μM. All the experiments were performed with oocytes from at least 3 different oocyte batches. Statistical differences were assessed using T-test or Mann-Whitney U test (Fe 2+ -uptake by hDMT1 WT vs. non-injected oocytes or vs. compound); p > 0.05 = ns; p < 0.001 = ***.
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

View Article Online
HEK293 cells expressing the transporter, at pH 5.5 for hDMT1 and at pH 7.4 for hZIP8. Indeed, activity screening showed that bipyridine, a wellknown iron chelator, non-specifically inhibited both transporters. Similarly, the literature pyrazolyl-pyridine 3, our optimized pyrazolyl-pyrimidone 13, and to a lesser extent pyrazolyl-pyrimidone 5, all inhibited both transporters, although the inhibition by 3 was not consistently observed at higher concentrations, probably due to adsorption of the complex to the cell material and assay plate, resulting in apparent uptake. On the other hand, bis-isothiourea 1, which inhibits hDMT1 by competing in the iron binding site, only inhibited hDMT1 and had no measurable activity on hZIP8 (Fig. 5a). A precise determination of the inhibitory effects with pyrazolyl-pyrimidone 13 and with the known iron chelator bipyridine showed that these compounds had comparable IC 50 values on both transporters (13: IC 50 ∼ 1.1 μM, bipyridine: IC 50 ∼ 6.5 μM, Fig. 5b/c).

Mechanistic model
We interpret the non-selective inhibition of bipyridine and 13 on both hDMT1 and hZIP8 as an indication that inhibition is caused by chelation of iron without specific interactions with the transporters. Metal chelation most likely involves a fivemembered ring chelate with the two cyclic nitrogen atoms within the N:NĲ:C)-C:N substructures, similar to the wellknown chelation by bipyridine forming a tight 3:1 complex with Fe 2+ . Acid-base titration of 3, 5 and 13 shows in each case two ionizable groups (3: pK a 1 = 3.3, pK a 2 = 6.7; 5: pK a 1 = 3.6, pK a 2 = 6.5; 13: pK a 1 = 3.3, pK a 2 = 7.0, Fig. 6a). The transition at neutral pH (pK a 2) can be attributed to the pyrazole hydroxyl group for 3 and to the pyrimidone NH group for 5 and 13, resulting in a transition from the neutral form below pH 7 to a mono-anionic form above pH 7 (Fig. 6a). Both forms possess an N:NĲ:C)-C:N substructure and should have metal chelating abilities, the neutral form existing at pH 5.5 where hDMT1 transport is measured, and the anionic form existing at pH 7.4 where hZIP8 transport is measured (Fig. 6b).
However, while all pyrazolyl-pyrimidones and pyrazolylpyridines tested here should have comparable metal chelating abilities, only a few of them actually showed activity in the iron uptake assays with hDMT1 and hZIP8. We therefore postulate that inhibition only occurs either if metal complexation is tight, as for the reference chelator bipyridine, or if metal complexation results in the formation of an insoluble complex, as observed using radioactive 55 Fe 2+ with 3 and 13. Note that, while iron uptake into hDMT1 expressing cells was partially reduced with the literature inhibitor 3, we did not observe inhibition at higher concentrations (Table 1), probably because complexed iron was in this case absorbed on the cell material and assay plate, and resulted in apparent iron uptake. This is in line with the Fe 2+ precipitation experiments at pH 5.5, where we only observe high precipitation of iron in the presence of compound 3. The lack of inhibition by most of the compounds tested might therefore either reflect partial iron chelation or formation of an iron complex that partially absorbs on the cells, resulting in apparent iron uptake and no net inhibition.

Pyrazolyl-pyrimidones and pyrazolyl-pyridines in chemical space
To better understand our compound series, we surveyed the ChEMBL database, which lists 1.82 million bioactive molecules and their annotated targets, 23 and found that 0.86% of these (15 585 molecules) contain a potentially metal chelating 2,2′-diazabiaryl group, including 1717 pyrazolylpyridines related to 3 and 169 pyrazolyl-pyrimidones related to 5 and 13 and annotated with bioactivities (Table 2). Among the 9.2 million screening compounds available in stock from commercial providers as listed in the ZINC database, 24 we similarly found 0.87% (80 153 molecules) 2,2′-diazabiaryls, however with relatively fewer pyrazolyl-pyridines (4673 Fig. 5 Cross reactivity of hDMT1 inhibitors over the functional activity of the divalent metal transporter hZIP8. (a) 55 Fe 2+ uptake (1 μM) by HEK293T cells (source: American Type Culture Collection, catalog no. CRL-3216) transiently transfected with the empty vector (Ev), hDMT1 or hZIP8. Uptake assay was performed at the optimal pH for the functional activity of each transporter (pH 5.5 and pH 7.4 respectively). The indicated compounds (10 μM) were pre-incubated for 5 min, and then, for 15 min with both compound and 55 Fe 2+ . Measured 55 Fe 2+ was corrected by subtracting the background iron uptake measured in non-transfected cells. The data was normalized to the Fe 2+ uptake measured in the presence of DMSO (0.1%) and represented as mean ± SD (N = 5-16; obtained from 2 independent experiments). IC 50 determination for pyrazolyl-pyrimidone 13 (b) and bipyridine (c) in HEK293T cells transiently transfected with hDMT1 (left panels) or hZIP8 (right panels). 55 Fe 2+ uptake (1 μM) was measured in the presence of the indicated compound concentration ranges, and the measurements were corrected by subtracting the background iron uptake measured in non-transfected cells. IC 50 values were calculated by fitting the data to a sigmoidal 4-parameter equation. Representative experiments are depicted, while IC 50 values were calculated as mean ± SD of 3 different experiments. molecules) but more abundant pyrazolyl-pyrimidones (1960 molecules), which suggests that such 2,2′-diazabiraryl compounds are relatively easy to synthesize and drug-like. The abundance of 2,2′-diazabiaryls was significantly lower in theoretically enumerated chemical space databases such as FDB17, 25 GDBMedChem 26 and GDBChEMBL, 27 reflecting the fact that these databases contain a much higher fraction of non-aromatic molecules.
We recently developed interactive chemical space visualization tools that provide useful insights into polypharmacology and structure-activity relationships. 28,29 To obtain a closer insight into the selected 2,2′-diazabiaryls, we created a tree-map (TMAP) 30 representing pyrazolylpyridines and pyrazolyl-pyrimidones from ChEMBL (Fig. 6c). The TMAP can be interactively analyzed in-browser by loading its layout using the program Faerun. 31 In this map each compound is represented by a point that can be colorcoded according to selected properties, and the molecular structure is shown upon mouse-over by the program Smilesdrawer. 32 The compounds are connected in a tree topology according to their structural similarities as measured by the MHFP6 molecular fingerprint. 33 The TMAP  illustrates the diversity of pyrazolyl-pyrimidones and pyrazolyl-pyridines in ChEMBL. The pyrazolyl-pyrimidones investigated here belong to the more drug-like molecules in the series, as can be appreciated by color-coding for druglikeness using the quantitative estimate of drug-likeness (QED). 34 The activities reported for these molecules in ChEMBL cover a diversity of target types, spanning from transporters and ion channels to enzymes and uncharacterized targets. We found 45 2,2′-diazabiaryl including 18 pyrazolylpyrimidones closely related to our compounds and reported with transporter activity in ChEMBL (Table S2 †). The ChEMBL record indicates, among various targets, inhibition of glucose and hexose transporters identified in antiparasitic screening campaigns against Leishmania mexicana, tuberculosis, and Plasmodium falciparum. 35 However, the screens performed were not transport assays but simple cytotoxicity tests. Considering that metal chelation is known to be a mechanism of toxicity against these parasites, 36,37 one can speculate that the reported transporter inhibitory activity of these pyrazolyl-pyrimidones might in fact reflect toxicity as a result of metal chelation. Further activities reported for pyrazolyl-pyrimidones and pyrazolyl-pyridines might similarly be the consequence of metal chelation rather than the attributed activity, including a recently reported activity of 3 and 5 against cancer cells. 38 Note that potentially metal chelating 2,2′-diazabiaryls such as pyrazolyl-pyrimidones are not listed in PAINS (pan-assay interference compounds). [39][40][41] Indeed a systematic checking of our compounds with the PAINS filter from RDkit does not flag any molecule in the series.

Conclusion
To date the pharmacology of divalent metal transporters is extremely limited. In view of the increasing importance of SLCs as drug targets, we investigated the inhibition of hDMT1 mediated iron transport by two previously reported inhibitors, namely pyrazolyl-pyridine 3 and pyrazolylpyrimidone 5. Optimization of 5 uncovered analog 13 showing potent inhibition of hDMT1 (IC 50 = 1.1 μM). However, this inhibitor did not modulate hDMT1-induced currents in Xenopus oocytes and exerted cross-inhibition of the unrelated iron transporter hZIP8.
Our combined experiments suggest that this non-selective inhibition reflects chelation of Fe 2+ in the assay medium by these molecules to form an insoluble precipitate. Although metal chelation is often exploited as a mechanism of drug action, 10-12 metal chelation is in principle an assay interference. This effect might occur across thousands of bioactive molecules reported in ChEMBL that share the same potentially chelating substructure and explain some of their reported bioactivities. Pyrazolyl-pyrimidones and related potentially metal chelating 2,2′-diazabiaryls are not considered in the list of pan-assay interference compounds (PAINS), however these should be checked whenever screening SLC targets.

Conflicts of interest
There are no conflicts to declare.