Mechanistic basis of the inhibition of SLC11/NRAMP-mediated metal ion transport by bis-isothiourea substituted compounds

In humans, the divalent metal ion transporter-1 (DMT1) mediates the transport of ferrous iron across the apical membrane of enterocytes. Hence, its inhibition could be beneficial for the treatment of iron overload disorders. Here we characterize the interaction of aromatic bis-isothiourea-substituted compounds with human DMT1 and its prokaryotic homologue EcoDMT. Both transporters are inhibited by a common competitive mechanism with potencies in the low micromolar range. The crystal structure of EcoDMT in complex with a brominated derivative defines the binding of the inhibitor to an extracellular pocket of the transporter in direct contact with residues of the metal ion coordination site, thereby interfering with substrate loading and locking the transporter in its outward-facing state. Mutagenesis and structure-activity relationships further support the observed interaction mode and reveal species-dependent differences between pro- and eukaryotic transporters. Together, our data provide the first detailed mechanistic insight into the pharmacology of SLC11/NRAMP transporters.


Introduction
Hereditary hemochromatosis (HH) is a multigenic iron overload disorder that results from the excessive absorption of iron in the intestine (Pietrangelo, 2010;Yen et al., 2006). In the absence of a regulated mechanism for its excretion, excessive iron can lead to significant tissue damage in the heart, liver, endocrine glands and other organs (Pietrangelo, 2010;Yen et al., 2006). The most prevalent form of HH is associated with the upregulation of the iron transport protein DMT1 (or SLC11A2) (Byrnes et al., 2002;Fleming et al., 1999;Rolfs et al., 2002;Stuart, 2003), which facilitates the uptake of ferrous iron (Fe 2+ ) across the apical membrane of enterocytes and whose expression is regulated on a transcriptional level (Fleming et al., 1997;Gunshin et al., 1997;Shawki et al., 2012). The current strategy to treat hemochromatosis is phlebotomy, which can have unwanted side-effects and which is not an option in cases of secondary hemochromatosis, such as thalassemia, since patients in this case are also anemic (Brissot et al., 2011;Gattermann, 2009). A potential alternative strategy to counteract excessive iron uptake would be the interference of transport by inhibition of DMT1 (Crielaard et al., 2017). Due to the accessibility of the transporter from the apical side, inhibition could proceed from the intestinal lumen by compounds that would not have to cross the membrane. substance class. These include five compounds carrying two isothiourea moieties for which we have varied the aromatic scaffolds (i.e. a brominated dibenzofuran and a single phenyl ring with different substituents) to investigate the influence of their respective size and geometry on inhibition ( Figure 1A, Appendix 1). For simplicity, we termed the tri-methyl and tri-ethyl substituted benzyl bis-isothiourea compounds TMBIT and TEBIT, respectively, and the dibenzofuran-based compound Br-DBFIT. Br-DBFIT, TMBIT and its derivatives were previously described as inhibitors of DMT1 . To ease the identification of benzyl bis-isothiourea compounds in inhibitor complexes by X-ray crystallography, we have also synthesized the brominated derivatives Br-BIT and oBr-BIT. Additionally, we have synthesized two variants of the inhibitor oBr-BIT where we have replaced one or both isothiourea moieties by bulkier thio-2-imidazoline groups. All molecules are water-soluble and thus poorly membrane-permeable with both basic isothiourea groups being    predominantly charged under physiological conditions (pKa = 8.5-9.5 as measured in a titration of TMBIT and Br-BIT, Figure 1-figure supplement 1A). We first tested the activity of all compounds on human DMT1 (hDMT1) by measuring radioactive 55 Fe 2+ transport into HEK293 cells stably expressing the protein. When assayed at a free Fe 2+ concentration of 1 mM, all compounds inhibit metal ion uptake in a dose-dependent manner with IC 50 values in the micromolar range ( Figure 1B, Figure 1-figure supplement 2). The most potent compounds TEBIT and TMBIT display IC 50 values of 0.27 mM and 0.35 mM, respectively, latter being in close quantitative agreement with a previous measurement using a calcein-based fluorescence assay (IC 50 = 0.29 mM) ( Figure 1B, Figure 1-figure supplement 2B,C) . In comparison, the larger values of Br-BIT (4.66 mM) and oBr-BIT (2.3 mM) indicate an equivalent interaction with somewhat lower affinity and the dibenzofuran compound Br-DBFIT (1.24 mM) is in our hands less potent than previously reported ( Figure 1B, Figure 1-figure supplement 2A,D and E) . To rule out that the observed activity would be due to chelation of divalent metal ions, we performed isothermal titration calorimetry measurements. Upon titrating MnCl 2 to either TMBIT or Br-BIT, we did not detect any pronounced response that would indicate specific binding ( Figure 1-figure supplement 1B), emphasizing that the inhibition of 55 Fe 2+ transport is caused by the specific interactions of either compound with hDMT1. We next investigated the role of the positively charged isothiourea groups for protein interactions by comparing the potency of oBr-BIT with its variants where one or two of the moieties were replaced. In case of the replacement of a single isothiourea group, we were able to measure a fourfold reduced potency of 8.13 mM whereas a much stronger reduction (IC 50 = 161 mM) was obtained for a compound where both isothioureas were modified (Figure 1-figure supplement 2F,G). Together these results underline the importance of the isothiourea moieties for specific protein interactions. To further characterize the mode of inhibition, we studied the effect of different extracellular inhibitor concentrations on the kinetics of iron transport ( Figure 1-figure supplement 3). In the absence of inhibitors, the transport rate at different 55 Fe 2+ concentrations can be fitted to a Michaelis-Menten equation with K M values of 2.6 mM to 4.4 mM and v max values of 2.7 to 6.1 pmol min À1 well À1 , which is in general agreement with previously reported values (Gunshin et al., 1997;Mackenzie et al., 2006;Pujol-Giménez et al., 2017). At increasing inhibitor concentrations, we observed in all tested cases a pronounced increase of the apparent K M whereas the apparent v max values decreased only slightly (Figure 1-figure supplement 3, Table 1). These results suggest that the compounds act by a predominant competitive mechanism. When fitting the data to a mixed enzyme inhibition model the resulting equilibrium constants are in the micromolar range with inhibitors binding with much higher affinity to the substrate-free transporter ( Table 1). Taken together, our data confirm the activity of aromatic isothiourea-based compounds as competitive inhibitors of hDMT1 with potencies in the low micromolar range. As all compounds are positively charged and thus membrane-impermeable, the binding site of the inhibitor is expected to be accessible from the extracellular side.

Functional characterization of the interaction with EcoDMT
After the characterization of hDMT1 inhibition, we have studied the properties of different bis-isothiourea compounds on the prokaryotic SLC11 homologue EcoDMT, which catalyzes H + -coupled Mn 2+ symport and whose structure was determined in an outward-facing conformation by X-ray crystallography (Ehrnstorfer et al., 2017). Due to the insufficient solubility of the dibenzofuran-based inhibitor Br-DBFIT and TEBIT for experiments with EcoDMT, we restricted this analysis to the benzyl bis-isothiourea compounds TMBIT, Br-BIT and oBr-BIT. To characterize EcoDMT-mediated transport, we have reconstituted the purified protein into liposomes and used a fluorescence-based in-vitro assay ( Figure 2-figure supplement 1A). In these proteoliposomes, EcoDMT is incorporated in inside-out and outside-out orientations at about equal ratios (Figure 2-figure supplement 1B). Concentration-dependent Mn 2+ uptake into proteoliposomes was monitored by the time-dependent quenching of the fluorophore calcein trapped inside the vesicles (Figure 2-figure supplement 1A) (Ehrnstorfer et al., 2017). In the absence of inhibitors, Mn 2+ transport by EcoDMT saturates at low micromolar concentrations (K M = 4.3 mM) (Figure 2-figure supplement 1C, Table 1). The addition of either benzyl bis-isothiourea compound decreases the kinetics of uptake in a dose-dependent manner thus suggesting that all tested compounds, when applied at micromolar concentrations to the outside of proteoliposomes, inhibit the transport activity of EcoDMT by binding to a saturable site of the protein (Figure 2A, Figure 2-figure supplement 1D,E). Since higher concentrations of TMBIT and oBr-BIT (i.e. >50 mM) did interfere with the assay, we restricted our quantitative characterization to Br-BIT, where we do not observe any interference at concentrations up to 200 mM. At high micromolar concentrations of Br-BIT, the decrease of transport activity approaches a maximum and even at 200 mM Br-BIT we could not detect complete inhibition. The saturation of the inhibition at high concentration results from the full occupancy of accessible binding sites, whereas the residual transport likely originates from transporters with inside-out orientation which do not expose the presumed inhibitor binding site to the external solution. The basal activity at high inhibitor concentration thus further demonstrates the sidedness of the inhibition and the membrane-impermeability of the compound. As for the inhibition of hDMT1, the K M values of transport increased at higher inhibitor concentrations, whereas v max did not show pronounced changes ( Figure 2B, Table 1). The K i value of 14.2 mM, representing the equilibrium dissociation constant to the substrate-free EcoDMT, is in the same range as the K i value of 3.6 mM obtained for hDMT1, reflecting the strong structural relationship between both proteins. Together, our results suggest that Br-BIT inhibits EcoDMT and hDMT1 by a common competitive mechanism.

Structural characterization of the inhibition of EcoDMT by Br-BIT
To investigate the structural basis for the inhibition of divalent metal ion transporters of the SLC11 family by benzyl bis-isothiourea-based compounds, we have characterized the interaction between the brominated analogs and EcoDMT by X-ray crystallography. In our experiments we exploited the anomalous scattering properties of the inhibitors to facilitate their localization in the complex. For that purpose, we have soaked crystals of EcoDMT with Br-BIT and oBr-BIT and collected multiple datasets at a wavelength corresponding to the anomalous absorption edge of bromine ( Table 2,  Table 3). Whereas we were unable to detect bromine in the anomalous maps of oBr-BIT containing crystals, the majority of datasets collected from crystals soaked with Br-BIT displayed a single strong peak in the anomalous difference density at equivalent positions, which aided the localization of the bound inhibitor ( Figure 3A,B, Figure 3-figure supplement 1A, Table 3). A detailed view of the complex defined in the 2Fo-Fc density at 3.8 Å is displayed in Figure 3A. In this structure, EcoDMT adopts the same substrate-free outward-facing conformation that has previously been observed in datasets of the protein in absence of the inhibitor ( Figure 3C,D) (Ehrnstorfer et al., 2017). In this conformation, a funnel-shaped aqueous pocket of the protein leads from the extracellular solution to the substrate binding site. The inhibitor is bound at the base of this pocket as defined by the anomalous difference density that constrains the position of the covalently bound Br-atom and by residual density in the 2Fo-Fc omit map that was calculated with phases from a model not containing the inhibitor ( Figure 3B). The fact that the Br-Atom of Br-BIT is located in the narrow apex of the pocket, whereas it would be placed in the wider part of the cavity in oBr-BIT might explain why we were unable to detect the binding position in the anomalous difference density of the latter compound. The omit map of the EcoDMT Br-BIT complex displays density for the aromatic ring and for the isothiourea group located close to the metal ion binding site (termed proximal isothiourea group), whereas the other group (the distal isothiourea group) is not defined in the electron density reflecting its increased conformational flexibility ( Figure 3B). In general, the shape of the binding pocket is complementary to the structure of the inhibitor but it is sufficiently wide in the long direction of the molecule to accommodate substitutions at the aromatic ring as found in the molecules TMBIT, TEBIT and in the larger dibenzofuran ring of Br-DBFIT ( Figure 3E). The aromatic group is stacked between a-helices 6 and 10 contacted by the side chains of residues Ala 231, Leu 410, Ala 409 and Leu 414. The close-by Asn 456 located on a11 might interact with the covalently attached Br atom of Br-BIT ( Figure 3F, Figure 3-figure supplement 1B). The proximal isothiourea group is located in a narrow pocket in interaction distance to the conserved Asp 51 and Asn 54 in the unwound part of aÀ1 and to Gln 407 on aÀ10, which were shown to contribute to the coordination of transported metal ions ( Figure 3E,F, Figure 3-figure supplement 1B) (Bozzi et al., 2019;Ehrnstorfer et al., 2014;Ehrnstorfer et al., 2017;Pujol-Giménez et al., 2017). The distal isothiourea group is located in the wider entrance of the cavity and might thus adopt different conformations, which is consistent with its undefined position in the electron density ( Figure 3B). In one conformation, this group approaches residues Ser 459 and Gln 463, both located on a11. Besides the direct ionic interactions of the proximal isothiourea group with Asp 51, the positive charge of both groups would be additionally stabilized by the negative electrostatics of the pocket that is conferred by an excess of acidic residues (Figure 3-figure supplement 1C). The observed binding position and the assumed interaction of the inhibitor with the metal ion binding site is also compatible with the observed competitive nature of the inhibition. The high sequence similarity between bacterial and human orthologs (i.e. 52% similar and 29% identical residues between EcoDMT and hDMT1) facilitates the construction of a homology model of human DMT1 (Figure 3-figure supplement 2A,B), which permits a glimpse of potential interactions of the inhibitor with the human transporter. As this model does not contain any insertions or deletions in the binding region, we expect a similar-shaped outward-facing cavity binding the inhibitor in hDMT1 as observed for EcoDMT ( Figure 3G, Figure 3-figure supplement 2B). The conservation is particularly high for a-helices 1 and 6 constituting the metal ion coordination site, but differences are observed for pocket-lining residues located on a-helices 10 and 11: While the corresponding residues Leu 414 in EcoDMT and Leu 479 in hDMT1 (both located on a10) seal the bottom of the binding cavity in both proteins, the hydrophobic character of Leu 410 and Ala 409 in EcoDMT, which contact one face of the aromatic ring is altered by the polar sidechains of Gln 475 and Ser 476 in hDMT1 ( Figure Figure 3F,G,H). Nevertheless, since both residues are located in the wider part of the binding pocket, it is justified to assume a similar general binding mode of the inhibitor in bacterial and human orthologues. As for EcoDMT, we expect that the strongly negative electrostatic potential within the binding pocket of hDMT1 favors the binding of the positively charged inhibitor (Figure 3-figure supplement 2C). Taken together, our structural data thus provide a detailed view of the molecular basis of the interaction of benzyl bis-isothioureabased inhibitors with divalent metal ion transporters of the SLC11/NRAMP family.

Functional characterization of inhibitor binding-site mutants of EcoDMT
To further characterize the binding of Br-BIT to EcoDMT, we have studied the effect of mutations of putative contact residues identified in the structure on inhibition ( Figure 4A). Although the described results emphasize the importance of interactions of the isothiourea group with the metal-   coordination site, these cannot be probed with the applied transport assays as mutations of coordinating resides interfere with ion uptake. We have thus employed isothermal titration calorimetry (ITC) to directly measure the effect of a metal-binding site mutant in EcoDMT on inhibitor binding.  In ITC experiments, we find two signals in the thermograms in response to the titration of the inhibitor to the WT protein. A weak endothermic contribution, which saturates at low mM concentrations (K D = 34.5 ± 5.0 mM) can be attributed to the loading of the inhibitor binding site and an exothermic signal saturating with an affinity in the mM range to a potential non-specific interaction with the protein ( Figure 4B, Figure 4-figure supplement 1A,B). To characterize the observed interaction between the positively charged isothiourea group and the negatively charged Asp 51 of the metalbinding site, we have expressed and purified the mutant D51A and measured inhibitor binding. Whereas the low-affinity signal in the thermograms appears unaltered, the high affinity component is absent, as expected if the mutant has removed an important interaction which interferes with inhibitor binding ( Figure 4C, Figure 4-figure supplement 1A,B). Thus, despite the weak signal originating from the low enthalpic contribution to binding, our titration calorimetry experiments indicate a direct interaction of the isothiourea group with the metal binding site as expected for a competitive inhibitor.
To probe the role of other residues of EcoDMT in the vicinity of the bound inhibitor, we have characterized the effect of alterations of three hydrophilic residues on a-helix 11 on the inhibition of Mn 2+ transport. Based on our structures, we suspected Gln 463 and Ser 459 to interact with the distal isothiourea group and Asn 456 with the bromine atom on the aromatic ring of Br-BIT ( Figure 4A). The three constructs, the single mutants N456A and N456L and the triple mutant N456A/S459A/Q463A transport Mn 2+ with similar kinetics as WT (  Table 1). In light of the small difference in K i compared to WT, our data excludes a large energetic contribution of residues on a11 to inhibitor binding, consistent with the assumed mobility of the distal isothiourea group that is manifested in the lack of electron density of the group in the structures of EcoDMT Br-BIT complexes.

Functional characterization of inhibitor binding-site mutants of hDMT1
To characterize the role of residues in the predicted inhibitor binding pocket of human DMT1, we have generated several point mutants and investigated the effect of these mutations on the interaction with different inhibitors. Due to the strong negative impact of alterations of the metal ion coordination site on transport, mutagenesis was restricted to residues lining the remainder of the binding pocket. The investigated positions encompassed residues on aÀ6 (Ala 291), aÀ10 (Gln 475, Ser 476 and Leu 479), and aÀ11 (Asn 520, Phe 523 and Tyr 527) ( Figures 3H and 5A). In our experiments we wanted to target interactions of protein residues with the aromatic ring in the narrow part of the binding pocket by either shortening the side-chains in the mutants A291G and Q475A, or by increasing their size in the mutants A291V, Q475F, S476V and L479F. In the orthogonal direction, the binding pocket is wider and would on one side be delimited by resides located on aÀ11 ( Figures 3G and 5A). Based on our model, we suspected the aromatic side chains of Phe 523 and Tyr 527 to be located in proximity to the distal isothiourea groups of TMBIT, TEBIT and Br-BIT or to the second phenyl-ring in the case of the dibenzofuran-based compound Br-DBFIT and Asn 520 in interaction distance with the aromatic ring harboring the proximal isothiourea group in all compounds ( Figures 3H and 5D). To probe these potential interactions, we have truncated the aromatic side chains in the mutants F523A and Y527A and generated a nearly isosteric hydrophobic substitution in the mutant N520L and subsequently studied the 55 Fe 2+ uptake properties of HEK293 cells transiently transfected with DNA coding for the respective constructs. Transport is similar to WT in case of the mutants S476V, F523A and Y527A, reduced in the mutants Q475A and N520L and undetectable in the mutants A291G, A291V, Q475F and L479F ( Figure 5B, Figure 5-figure supplement  1A). Mutations that render hDMT1 inactive, most likely interfere with structural rearrangements during ion transport, as judged by the tight packing of the respective region in the inward-facing structures of SLC11 transporters (Bozzi et al., 2016b;Bozzi et al., 2019;Ehrnstorfer et al., 2014). Inhibition experiments on hDMT1 were carried out with Br-BIT used for crystallization, the more potent inhibitors TMBIT and TEBIT and the dibenzofuran-based compound Br-DBFIT to explore the influence of the aromatic scaffold and the geometric relationship between the two isothiourea groups on interactions. Similar to WT, the addition of either compound at equivalent concentrations decreases uptake in the mutants Y527A and Q475A both located towards the extracellular entrance to the binding pocket ( Figures 3H and 5A,C and Figure 5-figure supplement 1B-D) thus suggesting that interactions with these residues do not strongly contribute to inhibitor binding. Conversely, the compounds had much smaller effects on the transport activity of cells expressing the mutants S476V, N520L and F523A located deeper in the binding pocket ( Figures 3H and 5C,D, Figure 5figure supplement 1B-C) thus suggesting that in these cases, the mutations affected inhibitor interactions. To further characterize the inhibitory properties of the investigated compounds, we have measured uptake at different inhibitor concentrations and found a strong reduction in potency in most cases ( Figure 5E, Table 4). Whereas the effect is uniform in the mutant S476V for all investigated compounds, the mutants N520L, and F523A showed a decreased potency of inhibition for the related molecules Br-BIT, TMBIT and TEBIT but only a slight reduction for Br-DBFIT ( Figure 5E Table 4) indicating that residues on aÀ11 might form distinct interactions with different inhibitor classes. This is consistent with the wide dimensions of the pocket in that direction that allows for a geometry-dependent placement of the aromatic ring and the attached isothiourea moiety on the distal side ( Figure 5D). Taken together our results suggest an involvement of residues on aÀ10 and aÀ11 on inhibitor binding to hDMT1 although with variable specificity, consistent with the proposed general binding mode of the inhibitors, which constrain the binding of the first aromatic ring to position the proximal isothiourea group in interaction distance with the metal ion coordination site. Since equivalent mutations of a11 in EcoDMT had little impact on inhibition of Br-BIT, our results also point towards species-dependent energetic differences in inhibitor interactions on the distal side of the inhibitor binding pocket, which are reflected in the poor conservation of residues in a11 and the wide geometry of the pocket in the prokaryotic transporter. Despite the described species-dependent differences, our data is generally consistent with the notion that the characterized compounds inhibit both pro-and eukaryotic transporters by binding to equivalent regions.

Discussion
By combining chemical synthesis with X-ray crystallography and in vitro binding and transport assays on human DMT1 and its prokaryotic homologue EcoDMT, our study has revealed detailed insight into the inhibition of SLC11 transporters by aromatic bis-isothiourea-based compounds. These compounds inhibit pro-and eukaryotic family members by a predominant competitive mechanism by binding to an outward-facing aqueous cavity leading to the transition metal ion coordination site (Figures 1, 2, 3 and 6) which prevents substrate loading and the transition to an inward-open conformation of the transporter. We have shown that these compounds do not interact with the reactive transported substrate, which has hampered the identification of specific inhibitors in high-throughput screens (Figure 1-figure supplement 1B). We have also shown that these compounds are positively charged and thus poorly membrane permeable and most likely attracted and stabilized by the strong negative electrostatic potential in the outward-facing aqueous cavity (Figure 1-figure supplement 1A, Figure 3-figure supplements 1C and 2C). Our structural studies have identified the binding mode of the inhibitors at the base of the funnel-shaped cavity, with the aromatic group snugly fitting into the pocket, thereby positioning the isothiourea group into ideal interaction distance with the aspartate of the transition metal binding site ( Figures 3A,B and 6B). Although the characterization of the interaction to the metal ion binding site is experimentally challenging, since mutations at this site interfere with transport (Bozzi et al., 2019;Ehrnstorfer et al., 2014;Ehrnstorfer et al., 2017;Pujol-Giménez et al., 2017), it is supported by several observations: First, the interaction of the isothiourea group with the metal ion binding site is displayed in the electron density of the complex ( Figure 3A,B). Second, the low micromolecular binding affinity of the inhibitor to the prokaryotic transporter EcoDMT observed in titration calorimetry experiments vanishes in a mutant truncating the binding site aspartate ( Figure 4B,C, Figure 4-figure supplement 1A,B)). Third, the interaction underlies the observed competitive mechanism that is shared by all investigated isothiourea-based compounds containing different aromatic substituents (Figure 1, Figure 1figure supplement 3), and fourth it underlines the strong requirement for the isothiourea group for potent inhibition. Latter is illustrated by the inhibition of human DMT1 by compounds where either one or both isothiourea groups are modified, leading to moderately reduced potency in the first, and a strongly reduced binding affinity in the second compound ( Figure 1B, Figure 1-figure supplement 2E,F,G). In our proposed inhibition mechanism, the role of the aromatic group in each compound is to position the inhibitor at the base of the predominantly hydrophobic pocket in proximity to the binding site ( Figure 6B). This is supported by the fact that a mutation in hDMT1 that likely narrows the pocket in this direction (S476V) leads to a reduced potency of inhibition ( Figure 5C,E, Figure 5-figure supplement 1B-E, Table 4). In the orthogonal direction, the funnel-  shaped pocket is sufficiently wide to accommodate larger groups, which might undergo successively stronger interactions, which is illustrated by the increased potency of two compounds containing additional alkyl modification at the aromatic ring (as it is the case for TMBIT and TEBIT) ( Figures 1B and 6B; Figure 1-figure supplement 2B,C). This general mode of interaction might also explain the inhibition of isothiourea-based compounds with larger ring systems as it is the case for the dibenzofuran Br-DBFIT ( Figures 1B and 6B; Figure 1-figure supplement 2A) and related compounds characterized in a previous study . Whereas one isothiourea group strongly interacts with the metal ion binding site in both pro-and eukaryotic transporters, the opposite groups reside in the wider exit of the cavity in a region that is poorly conserved between different SLC11 homologues ( Figure 3E,G, Figure 3-figure supplement 2A). In EcoDMT it most likely undergoes no specific interactions with the protein and instead exhibits large conformational flexibility as supported by the absence of electron density for this group in the X-ray structure of the inhibitor complex and by the mostly unaltered potency in mutants of potentially interacting residues ( Figures 3A,B, 4 and 6B). In contrast, mutations of equivalent positions in human DMT1 show a more pronounced effect thus pointing towards stronger inhibitor interactions distal to the metal binding site compared to EcoDMT (Figures 5 and 6B). This is generally supported by the reduced potency of an asymmetric compound binding to human DMT1 where only one of the isothiourea groups was modified (Figure 1-figure supplement 2F). In this case the effect of the modification could be explained by a moderate decrease in the interaction at the distal side where interactions with the protein might be less specific and by the reduced entropy of binding of the asymmetric compound with the metal ion binding site, which demands interaction with the isothiourea group ( Figure 6). A strategy to increase the potency and selectivity of compounds towards human DMT1 could thus rely on the optimization of interactions at the distal side of the binding pocket by a systematic variation of aromatic scaffolds and attached polar groups. In summary we have provided the first detailed mechanistic insight into the pharmacology of transition metal transporters of the SLC11 family. Our results are relevant for potential therapeutic strategies inhibiting human DMT1, which could be beneficial in cases where excessive uptake of iron in the intestine leads to iron overload disorders as observed in hereditary or secondary hemochromatosis, and our study provides a framework that might aid the improvement of these compounds to optimize both their potency and specificity.

Chemical synthesis
The chemical synthesis of all compounds is described in Appendix 1.

Cell lines
Experiments using human cell lines were conducted with HEK293 cells either stably (ATCC-CRL-1573) or transiently (ATCC-CRL3216) over-expressing DsRED-hDMT1 constructs. The cell line stably over-expressing hDMT1 has been characterized previously (Montalbetti et al., 2014). Mycoplasma contamination was negative for both cell lines as tested with the LooKOut Mycoplasm PCR Detection Kit (Sigma-MP0035). All cells were grown in DMEM media (Invitrogen) supplemented with 10% FBS, 10 mM HEPES and 1 mM Na-pyruvate at 37˚C, 95% humidity and air containing 5% CO 2 . For cells stably over-expressing hDMT1, the media was additionally supplemented with 500 mg ml -1 geneticin (Life Technologies).

Iron uptake and inhibition assays for hDMT1
For uptake experiments, HEK293 cells were grown in clear bottom, white-well, poly-D-lysine coated 96 well plates (Corning). Cells stably over-expressing hDMT1 were seeded 24 hr before the experiment at a density of 50,000 cells/well and cells used for transient transfection were seeded at 30.000 cells/well for 48 hr prior to the experiment and transfected 24 hr before the experiment using Lipofectamine 2000 (Life technologies) as described in the manufacturer's protocol. Briefly, culture media was removed from the wells and the cells were washed three times with uptake buffer (140 mM NaCl, 2.5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 1.2 mM K 2 HPO 4 , 10 mM glucose, 5 mM HEPES, 5 mM MES, pH 7.4). After the wash, the cells were incubated for 15 min at room temperature (RT) with uptake solution containing the indicated amount of non-radioactive ferrous iron (Fe 2+ ), 100 mM Ascorbic acid and 0.5 mCi/ml radiolabeled 55 Fe 2+ (American Radiolabeled) dissolved in uptake buffer (pH 5.5). After incubation, uptake solution was removed from the wells, and the cells were washed three times in ice-cold uptake buffer (pH 7.5). Before quantification, a scintillation cocktail (Mycrosinth 20, PerkinElmer) was added to each well, and the cells were incubated during 1 hr at RT under constant agitation. Accumulated radioactivity was measured using a TopCount Microplate Scintillation Counter (PerkinElmer). Transport rates were quantified with: To assess their inhibitory effect, cells were incubated with the indicated compounds at the specified concentrations during 5 min at RT prior to the addition of the uptake solution. To determine the kinetic parameters for the Fe 2+ transport mediated by hDMT1 WT and point mutants, the influx rates at different iron concentrations were fitted to the Michaelis-Menten equation. For the determination of IC 50 values, influx rates for each inhibitor concentration were plotted and data was fitted to a 4-parameter sigmoidal curve. Plotted influx rates correspond to the mean of the indicated biological replicates, errors are s.d. Each experiment was performed in duplicates for transiently transfected cells with data obtained from at least two independent transfections or triplicates for stably overexpressed WT hDMT1.

Expression and purification of EcoDMT
EcoDMT WT and mutants were expressed in E. coli MC1061 as C-terminally-tagged fusion proteins containing a 3C-protease cleavage site followed by a His 10 -tag. The tag was removed during purification unless specified otherwise. E. coli cells were grown in Terrific Broth (TB) medium supplemented with 100 mg ml -1 ampicillin, either by fermentation or in flasks. Cells were grown at 37˚C and the temperature was gradually decreased to 25˚C prior to induction. Protein expression was induced by addition of 0.0045% (w/v) L-arabinose at an OD 600 of~2.5 for fermenter cultures and~0.8 for cultures in flasks. For overnight expression the temperature was decreased to 18˚C and cells were subsequently harvested by centrifugation. All following protein purification steps were carried out at 4C . The cells were lysed in buffer A (20 mM HEPES, pH 7.5, and 150 mM NaCl) supplemented with 1 mg ml -1 (w/v) lysozyme and 20 mg ml -1 DNaseI using HPL6 high-pressure cell disruptor (MAXIMA-TOR). The lysate was subjected to a low-spin centrifugation (10,000 g for 20 min) and subsequently the membrane vesicles were harvested by ultracentrifugation (200,000 g for 1 hr). Membrane proteins were extracted by resuspending the vesicles in buffer A containing 10% (w/v) glycerol and 1-2% (w/v) of the specified detergents and subsequently the extract was cleared by centrifugation. The detergent n-decyl-b-D-maltopyranoside (DM, Anatrace) was used to purify proteins for reconstitution or crystallization experiments and n-dodecyl-b-D-maltopyranoside (DDM, Anatrace) for isothermal titration calorimetry (ITC). The extracted proteins were purified by immobilized metal affinity chromatography (IMAC). The GFP-His 10 tag was removed by addition of HRV-3C protease at a protein:protease molar ratio of 5:1 for 2 hr while dialyzing the sample against 20 mM HEPES, pH 7.5, 150 mM NaCl, 8.7% (w/v) glycerol, and 0.1% (w/v) DM or 0.04% (w/v) DDM. A second IMAC step was used to separate the GFP-His 10 tag and the protease from the cleaved protein. Subsequently, the purified membrane proteins were subjected to size exclusion chromatography on a Superdex S200 column (GE Healthcare) equilibrated in 10 to 20 mM HEPES, pH 7.5, 150 mM NaCl, and either 0.25% (w/v) DM or 0.04% (w/v) DDM. Peak fractions were used for reconstitution into liposomes, ITC and crystallization experiments. Purified samples of WT and mutant proteins were analyzed by SDS-PAGE and mass spectrometry.

X-ray structure determination
Crystals of EcoDMT were grown in 24-well plates in sitting drops at 4˚C by mixing 1 ml of protein (at a concentration of 7-10 mg ml -1 ) with 1 ml of reservoir solution consisting of 50 mM Tris-HCl pH 8.0-9.0 and 22-26% PEG 400 (v/v) and equilibrated against 500 ml of reservoir solution. Crystals grew within two weeks. For preparation of inhibitor complexes, crystals were soaked for several minutes with either Br-BIT or oBr-BIT. The two inhibitors were either added to the cryoprotection solutions at a final concentration of 5 mM or directly added as powder to the drops containing the crystals. For cryoprotection, the PEG 400 concentration was increased stepwise to 35% (v/v). All data sets were collected on frozen crystals on the X06SA or the X06DA beamline at the Swiss Light Source of the Paul Scherrer Institute on an EIGER X 16M or a PILATUS 6M detector (Dectris). Anomalous data were collected at the bromine absorption edge (0.92 Å ). Data were integrated and scaled with XDS (Kabsch, 2010) and further processed with CCP4 programs (Collaborative Computational Project, 1994). Structures were refined in Phenix (Adams et al., 2002) using the EcoDMT WT structure (PDB ID 5M87) as starting model. The model was modified in COOT (Emsley and Cowtan, 2004) and constraints for the refinement of the Br-BIT ligand were generated using the CCP4 program PRODRG (Schüttelkopf and van Aalten, 2004). Five percent of the reflections not used in refinement were used to calculate R free . The final refinement statistics is reported in Table 2. The coordinates of the EcoDMT-Br-BIT complex refined to data at 3.8 Å were deposited with the PDB under accession code 6TL2.

Modeling and Poisson-Boltzmann calculations
The electrostatic potential in the extracellular aqueous cavity harboring the inhibitor binding site was calculated by solving the linearized Poisson-Boltzmann equation in CHARMM (Brooks et al., 1983;Im et al., 1998) on a 150 Å Â150 Å Â 200 Å grid (1 Å grid spacing) followed by focusing on a 100 Å x 100 Å x 120 Å grid (0.5 Å grid spacing). Partial protein charges were derived from the CHARMM36 all-hydrogen atom force field. Hydrogen positions were generated in CHARMM, histidines were protonated. The protein was assigned a dielectric constant () of 2. Its transmembrane region was embedded in a 30 Å -thick slab ( = 2) representing the hydrophobic core of the membrane and two adjacent 10 Å -thick regions ( = 30) representing the headgroups. The membrane region contained a 38 Å -high and 22 Å -wide aqueous cylinder ( = 80) covering the extracellular aqueous cavity and was surrounded by an aqueous environment ( = 80). Calculations were carried out in the absence of monovalent mobile ions in the aqueous regions. The homology model of human DMT1 was prepared with the SWISS-MODEL homology modeling server (Biasini et al., 2014).

Reconstitution of EcoDMT into liposomes
EcoDMT WT and mutants were reconstituted using detergent destabilized liposomes according to Geertsma et al. (2008). The liposomes were formed using the synthetic phospholipids POPE and POPG (Avanti Polar lipids) at a w/w ratio of 3:1. The lipids where resuspended in 20 mM HEPES, pH 7.5, and 100 mM KCl after washing with diethylether and drying by exsiccation. Liposomes were subjected to three freeze-thaw cycles and extruded through a 400 nm polycarbonate filter (Avestin, LiposoFast-Basic) to form unilammellar vesicles. Triton X-100 was used to destabilize the liposomes and the reconstitutions were performed at a protein to lipid ratio of 1:100 (w/w) for transport assays and a protein to lipid ratio of 1:50 (w/w) to determine the orientation of the transporters in the liposomes. After detergent removal by the successive addition of Bio-Beads SM-2 (Bio-Rad) over a period of three days, proteoliposomes were harvested by centrifugation, resuspended in buffer containing 20 mM HEPES, pH 7.5, and 100 mM KCl and stored in liquid nitrogen. The orientation of the transporters in proteoliposomes was determined using a reconstitution of EcoDMT-His 10 in which the C-terminally Histidine-tag preceded by a 3C protease cleavage site has not been cleaved prior to reconstitution. Initially, proteoliposomes (containing a total of 2 mg lipids) were extruded using a 400 nm polycarbonate filter to generate unilammellar vesicles and split in two equal aliquots. Purified 3C protease was subsequently added to the outside of one aliquot of the proteoliposmes and incubated for 2 hr at room temperature. The external 3C protease was removed by washing twice with 20 volumes of 20 mM HEPES, pH 7.5, and 100 mM KCl and the liposomes were harvested by centrifugation. After removal of the protease, the liposomes were dissolved by addition of DM at a detergent to lipid ration of 1.25:1 (w/w) with half of the samples incubated with 3C protease for 2 hr on ice. All 3C cleavage steps were performed with a large excess of protease to ensure completion of the reaction. Control liposomes not treated with 3C protease at the different steps were processed the same way. A sample of purified EcoDMT-His 10 was used as control to follow the removal of the His 10 -tag in a sample with unrestricted accessibility to the 3C cleavage site. The final samples were analyzed by SDS-PAGE.

Fluorescence-based Mn 2+ transport and inhibition assays
Proteoliposomes for the Mn 2+ transport and inhibition assays were obtained by resuspension of vesicles in buffer B containing 20 mM HEPES, pH 7.5, 100 mM KCl and 250 mM calcein (Invitrogen) and subjection to three freeze-thaw cycles followed by extrusion through a 400 nm filter. Proteoliposomes were harvested by centrifugation and washed twice with 20 volumes of buffer B without Calcein. The samples were subsequently diluted to 0.25 mg lipid ml -1 in buffer containing 20 mM HEPES, pH 7.5 and 100 mM NaCl and varying concentrations of TMBIT, Br-BIT or oBr-BIT. Subsequently, 100 ml aliquots were placed in a black 96-well plate and after stabilization of the fluorescence signal, valinomycin (at a final concentration of 100 nM) and MnCl 2 were added to start the assay. Uptake of Mn 2+ into liposomes was recorded by measuring the fluorescence change in a fluorimeter (Tecan Infinite M1000, l ex =492 nm/ l em =518 nm) in four-second intervals. As a positive control, Mn 2+ ions were equilibrated by addition of the ionophore calcimycin (at a final concentration of 100 nM) (Invitrogen), which acts as a Mn 2+ /H + exchanger at the end of the experiments. In presence of TMBIT or oBr-BIT at concentrations higher than 50 mM, the fluorescence signal after addition of calcimycin did not reach the same low level as observed in absence of inhibitors, which suggests an interference of the compounds with the activity of calcimycin at high concentrations. Initial transport rates (DF min -1 ) were obtained by performing a linear regression of transport data obtained between 60 and 120 s after addition of valinomycin and MnCl 2 and fitted to a Michaelis-Menten equation. Kinetic data of WT and all mutants described in this study was measured in at least three independent experiments.

Analysis of kinetic data
Kinetic data was fitted to a mixed enzyme inhibition model outlined below (Scheme 1) (Copeland, 2005) with GraphPad Prism (GraphPad Software, San Diego, California USA, www.graphpad.com):

Scheme 1. Mixed enzyme inhibition model.
This model assumes that the inhibitor (I) binds to the substrate free transporter (T) and to the transporter-substrate complex (TÁM 2+ ) with equilibrium constants K i and K ii , respectively. Both equilibrium constants can be obtained by a non-linear regression to Equation 2 For high values of a, the inhibitor preferentially binds to the substrate-free transporter and Equation 2 approaches a model for competitive inhibition. The resulting equilibrium constants obtained for hDMT1 using a radioactive 55 Fe 2+ transport assay and for EcoDMT1 using an in vitro proteoliposome-based assay are summarized in Table 1.

Isothermal titration calorimetry
Isothermal titration calorimetry experiments were performed with a MicroCal ITC200 system (GE Healthcare). The titrations of MnCl 2 to TMBIT and Br-BIT were performed at 25˚C in 20 mM HEPES, pH 7.5, and 100 mM KCl. The syringe was filled with 5 mM MnCl 2 and sequential aliquots of 2 ml were added to the sample cell filled with 0.4 mM TMBIT, Br-BIT, Ethylenediaminetetraacetic acid (EDTA) or buffer. The titrations of Br-BIT to purified EcoDMT were performed at 6˚C in 20 mM HEPES, pH 7.5, 150 mM and 0.04% (w/v) DDM. The syringe was filled with 1.8 mM or 2.5 mM Br-BIT and sequential aliquots of 1.5-2 ml were added to the sample cell filled with~50 mM or~180 mM EcoDMT WT, the mutant D51A or buffer. Data were analyzed using the Origin ITC analysis package and the MicroCal ITC program Concat and errors on the reported K D values represent fitting errors. The data were fit using models assuming one or two sets of binding sites. In case of 2.5 mM Br-BIT in the syringe and~180 mM EcoDMT in the cell, mainly the high affinity step saturating in the low micromolar range is titrated. Therefore, the low affinity transition can be ignored and the resulting reaction enthalpies were y-translated to zero to enable data analysis using a model assuming a single set of binding sites. For each protein, similar results were obtained for at least two experiments from independent protein preparations.

Data availability
The coordinates and structure factors of the EcoDMT-Br-BIT complex have been deposited in the Protein Data Bank with the accession code 6TL2.