T-REx™/FPN-V5 cells were generated as described (Ross et al., 2012). T-REx™/FPN-V5 cells were engineered with a ß-lactamase (BLA) reporter gene joined to the 5′UTR [containing one copy of the iron response element (IRE)] of ferritin to generate the T-REx™/FPN-V5/IRE-BLA cell line. The nucleotide sequence of the ferritin IRE is provided in the international patent application WO 2009/094551 A1 (Arvedson et al., 2009). IRE-BLA sequences were cloned into pENTR1A and stably transfected into T-REx™/FPN-V5 cells. 293 T-REx™/FPN-V5, T47D, and UT-7/Epo cell lines were cultured as described (Ross et al., 2012).

Rabbit anti-Cyclophilin B antibody (Abcam, #ab16045) was used at 0.5 μg/ml as a loading control for immunoblots. Rabbit anti-V5 antibody (Abcam #9116) was used at 1 μg/ml as a control for FPN from engineered cell lines in immunoblots. Tested anti-FPN antibodies include Alpha Diagnostics (MTP11-A, used at 5 μg/ml) and LifeSpan Bioscience (LS-B1836S, used at 1.5 μg/ml) for immunoblotting. Anti-CD68 (Santa Cruz #sc-9139), anti-GFAP (Biocare #040) rabbit polyclonal antibodies were used for staining of macrophages and astrocytes, respectively, followed by goat-anti-rabbit Alexa-fluor® 594 (Invitrogen/Thermo Fisher #A-11012).

Mouse anti-human FPN antibodies were generated in C57BL/6 mice (Charles River Laboratories) by immunizing with either membranes from human FPN-expressing Chinese Hamster Ovary (CHO) cells or with recombinant adenoviral vector carrying human FPN. Mouse anti-human FPN antibody 31A5 was generated by immunizing C57BL/6 mice (Charles River Laboratory) with an adenovirus expressing human FPN and boosted with FPN DNA as described in the published international patent application WO 2009/094551 A1 (Arvedson et al., 2009). Fully human antibodies were generated in XenoMouse™ (Mendez et al., 1997) by immunizing mice with membranes from HEK293-6E cells expressing FPN using pTT5 plasmid (cells and plasmid licensed from the National Research Council of Canada Biotechnology Research Institute). Hybridomas were generated using standard techniques and clones expressing antibodies that bound the extracellular region of FPN were selected by screening hybridoma supernatants against live HEK 293T cells expressing inducible human FPN (T-REx™/Fpn-V5). Anti-mouse FPN antibodies were generated by immunizing rats with membranes from mouse FPN-expressing CHO cells. Mouse FPN pTT5 constructs were designed with and without PADRE T cell epitope tags and the protein transiently expressed in HEK293 cells. Expression cell pools were harvested and further processed into membrane preparations for use as antigen. Groups of Brown Norway rats (Charles River) were immunized 3–5 times with 100–200 μg per immunization of mouse FPN-PADRE membranes emulsified in the following adjuvants: complete Freund's adjuvant (CFA, Sigma), incomplete Freund's adjuvant (IFA, Sigma), Alum (Thermo Fisher), CpG (Amgen). Serum titers were analyzed by flow cytometry (BD) on CHO cells expressing mouse FPN and mock-transfected CHO cells. Spleens from rats that showed the best serum titer response were harvested for hybridoma generation by electrofusion. Due to contaminating host cell immune responses arising from immunizing with mouse FPN membrane preparations, bacolovirus mouse FPN-V5 constructs were expressed in TNi insect cells. Mouse FPN-expressing TNi cells in conjunction with goat anti-rat IgG, Fc specific-647 secondary antibodies (Jackson ImmunoResearch) were used to screen hybridoma supernatants, immune rat serum, and purified IgGs on the FMAT platform (Applied Biosystems) as well as by flow cytometry (BD).

Peptides of 10 amino acids were synthesized using a MultiPep instrument (Intavis) directly on cellulose membranes (Intavis) using the peptide spot (PepSpot) array technique (Briant et al., 2009). Antibodies were screened against individual overlapping FPN peptides (10 amino acids in length over the entire length of the FPN sequence) immobilized on membranes in individual spots. The first spot contained a peptide with amino acids 1–10, the second spot contained a peptide with amino acids 2–11, followed by 3–12, and so on. This pattern was followed over the length of the ferroportin sequence to ensure the entire protein was represented. Membranes were soaked in methanol for 5 min, washed twice in Tris-buffered saline, 0.1% Tween 20 (TBST) buffer, and blocked with 5% non-fat dried milk (NFDM) in TBST. Antibodies were screened at 1 μg/ml in 5% NFDM for 1 h with shaking at room temperature. After washing four times in TBST, membranes were incubated with goat anti-human IgG-HRP (Pierce/Thermo Fisher) secondary antibody at 1:50,000 in 5% NFDM. After washing four times in TBST, spots were detected by enhanced chemiluminescence (ECL, Pierce/Thermo Fisher).

Cells were plated in 96-well Poly-D-Lysine coated plates (BD) at 50,000 cells/well and induced overnight with 10 ng/ml doxycycline (Clontech). Cell surface FPN and RhoG-hepcidin uptake assays were performed as described (Ross et al., 2012). Briefly, induction medium was replaced with assay medium (DMEM/10% dialyzed FBS (for T-REx™/FPN-V5 cells) or growth medium (for all other cell lines) containing hepcidin. Cells were incubated at 37°C at varying concentrations of hepcidin for varying times (generally 30–360 min). For cell surface FPN detection, hepcidin-containing medium was replaced with anti-FPN antibodies at concentrations listed in figure legends in cold assay medium, incubated for 1 h at 4°C, washed twice, and incubated for 1 h with secondary antibodies (goat anti-mouse-Alexa Fluor 647 or goat anti-human-Alexa Fluor 647, Invitrogen/Thermo Fisher) and Hoechst 33342 nuclear dye (Invitrogen/Thermo Fisher). After staining, cells were fixed with 3.7% formaldehyde solution (Sigma) for 10 min at room temperature and washed with phosphate-buffered saline (PBS, GIBCO). For total FPN staining, cells were fixed with 3.7% formaldehyde solution and permeabilized with 0.1% saponin (Calbiochem) and 1% bovine serum albumin (Sigma) in PBS at room temperature. FPN staining was completed as described above except that the incubations were performed at room temperature. For RhoG-hepcidin uptake assays, T-REx™/FPN-V5 cells were treated with 250 nM RhoG-Hepc for varying times (generally 30–360 min.) and fixed with 3.7% formaldehyde. Nuclei were stained with Hoechst dye. Cell surface or total FPN or intracellular hepcidin was quantitated with cellular imaging. Uninduced (non-FPN-expressing) cells served as a negative control. Images were captured from Arrayscan™ VTI (Thermo Fisher) equipped with Zeiss microscope optics and Hamamatsu ORCA®-ER CCD camera. Images were captured with 10X or 20X objectives and analyzed with the Spot Detector bioapplication. A minimum of 300 cells per well were analyzed for each condition.

Cells were lysed in 20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100 with protease and phosphatase inhibitors (NaF, Roche protease inhibitor, and Sigma phosphatase inhibitors I and II). Protein was quantitated using a BCA protein assay (Thermo Fisher). For FPN analysis, non-reduced lysates were heated at 37°C for 10 min prior to loading on a 4–20% Tris-Glycine gel (Novex) and transferred to nitrocellulose membranes. FPN was detected with anti-FPN antibodies at concentrations listed in Figure Legends. Human cyclophilin B (CypB, rabbit polyclonal antibody, Abcam) was used as a protein loading control. HRP-labeled, species-specific secondary antibodies were used for detection (Cell Signaling).

T-REx™/FPN-V5 cells were reverse transfected with 20 nM SLC40A1 (Invitrogen/Thermo Fisher #HSS121213), cyclophilin B (QIAGEN, custom), or All Stars negative control (QIAGEN #1027281) siRNAs using RNAiMAX transfection reagent (Invitrogen/Thermo Fisher) according to the manufacturer's protocol. Cells were induced to express FPN with 10 ng/ml doxycycline immediately before transfection. Lysates were made 48 h after transfection. Protein (20 mg) from each sample was tested by Western analysis as described.

All human specimens were collected under Institutional Review Board approval with appropriate informed consent. In all cases, materials obtained were surplus to standard clinical practice. Patient identity and PHI/identifying information were redacted from tissues and clinical data. Human tissue specimens were obtained from the following institutions: Asterand Bioscience, Detroit, MI and Zoion, Hawthorne, NY.

Formalin-fixed, paraffin-embedded sections of human tissue were stained with a mouse monoclonal anti-human FPN Ab 31A5. Tissue sections were stained via indirect immunofluorescence or with the ABC-peroxidase technique. Brightfield mages were collected with a Nikon Ni-U microscope and a DS-Fi2 camera. Fluorescent images were collected with a Nikon Eclipse 50i fluorescence microscope illuminated by an EXFO Excite 120 light source and acquired using NIS elements v3.0 software.

Human FPN probe: A 389 bp fragment of the human FPN gene, corresponding to nucleotides 1632–2020 (Genbank #AF226614.1), was cloned into the pCR4-TOPO plasmid vector (Thermo Fisher). The identity of the template was verified by sequencing. An antisense ³³P-labeled RNA probe was synthesized by in vitro transcription of the template with T3 RNA polymerase after linearization of the vector with Not I restriction enzyme. A ³³P-labeled sense probe was also generated from the same template using T7 RNA polymerase and Spe I restriction enzyme. All of the tissue used in the study was derived from archived blocks of immersion fixed, paraffin embedded material from which 5 μm sections were taken. A standard ISH protocol (Wilcox, 1993) was performed involving overnight hybridization at 60°C in a hybridization solution containing 1 × 10⁶ cpm of ³³P-labeled riboprobe per slide. To improve target detection, all tissue slides were subjected to a pretreatment by microwave heating to 100°C totaling 10 min in a citric acid buffer solution (Citra—Biogenex) prior to hybridization. After overnight hybridization all slides were subjected to RNase digestion followed by a series of SSC washes with the highest stringency of 0.1X SSC at 55°C for 30 min. The slides were coated with Kodak NTB emulsion and exposed for 3 weeks in the dark at 4°C, developed, and then counterstained with hematoxylin and eosin.

Human FPN cDNA was targeted at the ATG starting codon of the mouse FPN locus, and ended at the stop codon, keeping all of the 3′UTR of the mouse gene intact, and replacing the entire mouse FPN locus with human FPN cDNA. The FPN cDNA with Neo selection cassette inserted at the 3′ end of the FPN gene was flanked by homology arms. The floxed Neo cassette was removed by cre recombinase in 129Sv (agouti) embryonic stem (ES) cells. ES cell clones were karyotyped and microinjected into C57BL/6 blastocyst embryos. Chimeric (129Sv/C57BL/6) blastocysts were microinjected into C57BL/6 mice. Male 8-week old mature chimera (F0) were crossed with female C57BL/6 mice to obtain germline transmitted F1 heterozygotes. Only heterozygous mice were obtained.

T-REx™/FPN-V5 cells plated in 96-well Packard ViewPlates at 50,000 cells/well and induced overnight as described for the BLA screening assay. Cells were incubated with compound for 30 min, followed by two washes to remove compound before addition of 250 nM RhoG-hepcidin for 1 h. As a control, cells were treated identically except that compound was not removed before addition of RhoG-hepcidin. Cells were fixed with 4% formaldehyde (Thermo Fisher) and stained with Hoechst nuclear dye. RhoG-Hepc uptake per cell was measured by cellular imaging.

Induced and uninduced cells were treated ± ³H-quinoxaline at 0.33 μM, and incubated at 37°C for 30 min. Unbound quinoxaline was removed by centrifugation. Non-reduced clarified lysates were run on a 4–12% Bis-Tris gel. The gel was cut into 2 parts; one side was dried on a BioRad gel dryer and exposed to autoradiography film for 1 month. The other side was transferred to nitrocellulose membrane and FPN detected with antibody 38C8.

T-REx™/FPN-V5 cells were treated ± 0.25–1 μg/ml tunicamycin (Sigma-Aldrich) for 18 h, followed by treatment with 500 nM hepcidin for 5 h for blots or 3 h for a RhoG-hepcidin uptake assay.

Synthetic human hepcidin alanine mutants were synthesized on a Tetras asynchronous automated synthesizer (Creosalus Louisville, KY) using N^α-Fmoc/^tBu side chain protection with (1-Cyano-2-ethoxy-2-oxoethylidenaminooxy)dimethylamino-morpholino-carbenium hexafluorophosphate (COMU)/ N,N-Diisopropylethylamine (DIEA) (1:1.2) chemistry in dimethylformamide (DMF) with Fmoc deprotection employing 20% (v/v) piperdine in DMF. H-Thr(^tBu)-ChemMatrix resin ChemMatrix Quebec, Canada) was used at 0.10 mmol equivalent scale. The following Fmoc protected amino acids were used, Thr(^tBu), His(Boc), Phe, Pro, Ile, Cys(Trt), Gly, Arg(Pbf), Lys(ivDde), Lys(Boc), Met, and Asp(^tBu). 5 equivalents of activated amino acid were coupled to the deprotected amine for 45 min followed by 2 × 15 min Fmoc deprotection with 20% piperdine in DMF. Alanine mutant peptides were tested for activity in both BLA and ferroportin internalization assays.

Sulfonyl quinoxaline 1 is 2-[(4-chlorophenyl)sulfonyl]quinoxaline and is commercially available from Key Organics Ltd., Cornwall, UK (CAS #: 338394-53-9; www.keyorganics.net). Tritiated quinoxaline 2 was obtained from Moravek Inc., Brea, CA.

Peptide and small molecules (100 μM each) were incubated together in 20 mM acetate buffer, pH 7 for 1 h (iodoacetamide) or 12 h (compound 1) prior to analysis by mass spectrometry on a Bruker 7 T Apex IV Fourier transform ion cyclotron resonance (FTICR) mass spectrometer. Samples were introduced to the mass spectrometer by electrospray ionization.

Lysates from doxycycline-induced T-REx™/FPN-V5 cells (370 μg of protein) were immunoprecipitated with anti-V5 antibody (Abcam) and protein G Dynabeads (Thermo Fisher). Immunoprecipitates were treated with PNGase F (New England BioLabs) according to New England BioLabs protocol with the following modifications: (1) denaturation at 50°C for 30 min and (2) PNGase F treatment for 3 h at 37°C. Samples were electrophoresed on a 4–20% Tris-Glycine gel, transferred to nitrocellulose and probed with Abs 38C8 (20 μg/ml) and 38G6 (1 μg/ml).

GraphPad Prism® 6.07 was used for graph generation and analysis. Four-parameter variable slope non-linear regression was used for dose response curve fitting. Statistical significance was determined using an unpaired t-test and two-tailed P-values.

All biological and chemical materials and procedures were performed in strict compliance with Amgen Environmental Health and Safety rules.

Mice and rats were housed in groups at an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC), International accredited facility. Animals were cared for in accordance with the Guide for the Care and Use of Laboratory Animals, 8th Edition. All research protocols were reviewed and approved by the Amgen Institutional Animal Care and Use Committee. Mice and rats (Charles River Laboratories) female, 6–8 weeks of age, were housed in individual ventilated caging (IVC) system on an irradiated corncob bedding (Envigo Teklad 7097). Lighting in animal holding rooms was maintained on 12:12 h light:dark cycle, and the ambient temperature and humidity range was at 68–79 F and 30–70%, respectively. Animals had ad libitum access to irradiated pelleted feed² (Envigo Teklad Global Rodent Diet- soy protein free extruded 2020X) and reverse-osmosis (RO) chlorinated (0.3–0.5 ppm) water via an automatic watering system. Cages were changed biweekly inside an engineered cage changing station. For XenoMouse™: Research and all technical procedures performed on animals under this study were approved by the Animal Care Committee (ACC) for Amgen British Columbia. Animals were housed in a Canadian Council on Animal Care (CCAC) accredited facility and were cared for according to standards established by the CCAC and comply with institutional policies and guidelines. Animal experiments were executed in strict compliance with institutional guidelines and regulations.

Several previous studies have relied on expression of tagged FPN (e.g., GFP, MYC, FLAG) in engineered cells as the only means of detecting FPN, precluding the possibility of detecting endogenously-expressed protein. Tags such as MYC and FLAG encoded on the N- or C-termini of FPN also prevented detection in live cells, since both termini are intracellular (Liu et al., 2005; Rice et al., 2009; Wallace et al., 2010; Taniguchi et al., 2015). Anti-FPN antibodies are available from commercial sources, and some of these antibodies have been used in published studies to detect human FPN (Pinnix et al., 2010; Bonilla et al., 2012). We evaluated a number of these antibodies but we were unable to demonstrate that they detected human FPN (Supplementary Figure 1A), although one of these antibodies did detect mouse FPN (Supplementary Figure 1B). We generated anti-human FPN antibodies by immunizing mice (XenoMouse™ for fully human antibodies and C57BL/6 mice for mouse antibodies) with membranes from human FPN-expressing CHO cells. Hybridoma clones expressing antibodies that bound the extracellular region of FPN were selected by screening hybridoma supernatants against live human FPN-expressing 293T (T-REx™/FPN-V5) cells. The anti-human FPN antibodies discussed herein are 31A5, 45D8, 38C8, and 38G6 (Table 1). Anti-mouse FPN antibodies were generated by immunizing rats with membranes from mouse FPN-expressing CHO cells. Clones expressing antibodies that bound the extracellular region of mouse FPN were selected by screening hybridoma supernatants against live insect cells infected with a baculovirus-encoding mouse FPN. The anti-mouse FPN antibodies discussed herein are 1C7 and 1E11 (Table 1).

Characterization of the binding epitope recognized by the FPN antibodies revealed that three of the six anti-human and anti-mouse antibodies (31A5, 38G6, 1C7) recognized sequences in the fifth extracellular (EC) loop (Figure 1). Sequence alignment shows that the amino acid sequence in the fifth extracellular loop between transmembrane helices 9 and 10 differs between human, mice and rats, whereas the regions on either side of this loop, containing residues important for binding hepcidin (including C326) are 100% conserved (Supplementary Figure 2). The sequence diversity in loop five may explain why the antibodies were primarily directed to this region.

Both the anti-mouse and -human FPN antibodies were capable of detecting FPN in multiple formats including immunoblotting, immunofluorescence, immunohistochemistry and flow cytometry. Table 1 provides a description of the antibodies tested and the applications for which they are suitable. Most of the antibodies were interchangeable with respect to applications, with few exceptions as noted in Table 1. Analogous data sets were generated for most of the antibodies discussed in this manuscript, and choice of using one antibody over another was often based on availability at the time of study. To characterize the antibodies, we frequently used engineered V5-tagged FPN-transfected T-REx™-293 cells (T-REx™/FPN-V5), where FPN expression was inducible by doxycycline (hereafter “doxycycline-induced” is referred to as “induced”). Antibody specificity was demonstrated by characterization of antibody binding to (1) uninduced and induced cells, (2) siRNA-treated induced cells and (3) hepcidin-treated cells. Western analysis results for two anti-human FPN antibodies (38C8 and 31A5), along with control anti-V5 antibody are shown in Figure 2A. FPN comprises 571 amino acids and contains 3 N-linked glycosylation sequences. On a non-reducing SDS-PAGE gel, FPN migrates at ~65 kDa. Antibodies 38C8 and 31A5 detect a band of ~65 kDa in lysates from induced cells and in lysates from induced cells treated with non-silencing siRNA control. Neither antibody is reactive with lysates from uninduced cells or cells treated with FPN siRNA. Since FPN has a V5 tag in these cells, an anti-V5 antibody (Abcam) was used as a positive control, which also detected a band of ~65 kDa. Based on these results we concluded that the 65 kDa band was FPN. Cyclophilin B (22 kDa) was used as a control for loading and to verify knockdown (Figure 2A). In addition to detecting FPN by Western analysis, it was also possible to detect surface-expressed FPN by immunofluorescence. As shown for anti-human FPN antibody 45D8, cell-surface FPN was detected on induced cells, but not on uninduced cells or on induced cells treated with hepcidin (Figure 2B).

We identified cell lines that endogenously express human FPN, including the ductal breast epithelial cell line T47D and the erythropoietin-dependent erythroleukemia cell line UT-7/Epo. Surface FPN was detectable using the anti-human FPN antibodies described above and this expression could be increased by treatment with ferric citrate (not shown) and decreased in response to treatment with hepcidin (Figure 3A). Using these antibodies, it was possible to develop a quantitative cellular imaging assay that measured hepcidin-induced FPN internalization. Data from this assay indicated that T47D cells responded to hepcidin comparably to the engineered T-REx™/FPN-V5 cell line, with EC₅₀ values for FPN internalization of 13 nM and 11 nM, respectively, after 6 h of treatment (Figure 3B). Hepcidin-mediated FPN degradation in UT-7/Epo cells was corroborated by Western analysis where the FPN signal had nearly disappeared after 6 h (Figure 3C).

Endogenous human FPN expression was also detectable by immunohistochemistry (IHC), demonstrated using antibody 31A5. This antibody recognizes a sequence in human FPN that is not conserved in mouse FPN. To demonstrate specificity for human FPN, antibody 31A5 was tested on human spleen, mouse spleen and mouse spleen in which human FPN had been knocked in. As shown in Figure 4A, antibody 31A5 produced a robust signal in human spleen and in the mouse spleen expressing human FPN; however, it did not produce a signal in mouse spleen. Specificity was further confirmed in human breast tissue where FPN detection by IHC was corroborated by in situ hybridization (ISH, Figure 4B). FPN expression has been reported in the duodenum, reticuloendothelial macrophages and placenta of mice (Abboud and Haile, 2000; Donovan et al., 2005). We investigated FPN expression in a panel of human tissues. In the duodenum, FPN expression was confirmed on the basolateral surface of enterocytes (Figure 4C, Donovan et al., 2000). FPN expression was detected in Kupffer cells of the liver (Figure 4D) and in interstitial mononuclear cells from esophagus, stomach, intestine, kidney, prostate, cervix, and skin (not shown). FPN was also detected in syncytiotrophoblasts of the placenta (Figure 4E).

We also detected FPN in tissues for which expression has not been previously described. Antibody 31A5 detected expression in neuronal astrocytes (Figure 5A), and robust expression of FPN was detected in the cortical, but not medullary cells of the adrenal gland (Figure 5B). FPN expression in adrenal cortex was confirmed by ISH (Figure 5C), and was localized to the membrane (Figure 5D), suggesting its function as an iron exporter is preserved in this tissue. Expression in adrenal gland is corroborated by RNA data from the Genotype Tissue Expression (GTEx) Project database for RNA sequencing data where adrenal gland is the tissue with the highest mean FPN mRNA content of all tissues characterized (Figure 5E). It is unclear why FPN is so highly expressed in the adrenal cortex.

The anti-FPN antibodies provided a means of monitoring changes in FPN expression. To develop a trackable form of hepcidin, we initially labeled the N-terminus with Rhodamine green (RhoG) using linkers of different lengths; however, all N-labeled forms of hepcidin were inactive (not shown), corroborating previously-published work identifying the N-terminus as necessary for binding to FPN (Nemeth et al., 2006). We then initiated a campaign to synthesize hepcidin analogs that were amenable to labeling at positions other than the N-terminus. Hepcidin is a tightly-folded peptide containing four disulfide bonds that are conserved across species from fish to mammals (Shike et al., 2004; Jordan et al., 2009). We used alanine scanning to identify a derivatization point for the label by individually replacing all non-cysteine amino acids with alanine and testing them for bioactivity in an assay that monitors surface FPN expression (FPN internalization assay) and an iron response reporter assay. Briefly, the iron response assay detects intracellular iron accumulation through the expression of a reporter controlled by an iron response element (IRE) (Li et al., 2004). T-REx™/FPN-V5 cells were engineered with a β-lactamase (BLA) reporter gene downstream of the ferritin IRE, and BLA activity was detected by ratiometric fluorescence resonance energy transfer (FRET) upon addition of substrate. We found most substitutions resulted in reduced activity when compared to wild type (WT), with hydrophobic and aromatic residues in the N-terminal portion of the molecule having the most pronounced effect (I8, F4, I6, P5, F9, and H3). Substitutions at H15, K18, and T2 resulted in increased activity, while substitutions at residues R16, S17, and T25 were roughly equipotent to WT hepcidin (Table 2). Given the current model of the hepcidin/FPN interaction, where residues 16 and 17 are predicted to be pointing away from the complex (Taniguchi et al., 2015), it is not surprising that these residues could be labeled and still maintain the ability to bind FPN. We mapped the outcome of the alanine mutagenesis activity data on the crystal structure (Figure 6, Jordan et al., 2009) and selected S17 for derivatization as this residue is distal from the N-terminus which is reported to be necessary for FPN binding, and replacement of serine with RhoG labeled lysine (neutral charge) would not change the overall charge of the molecule. We tested linkers of different molecular weights and selected one of the lowest molecular weight linkers, comprising a single amino hexanoic acid (Ahx) molecule as analogs with longer linkers were inactive (data not shown). The final analog that we refer to as RhoG-hepcidin is [Lys17(N^ε-Rhodamine green amino hexanoyl)]hepcidin. We found that the labeled analog was similar in potency to WT unlabeled hepcidin when measuring FPN internalization by cellular imaging (Figure 7A), EC₅₀~30 nM in this assay. As expected, RhoG-hepcidin uptake was apparent in induced, but not uninduced cells (Figure 7B) indicating that interaction with FPN was required for hepcidin internalization. Using RhoG-hepcidin, it was possible to concurrently monitor hepcidin uptake and FPN internalization. Quantification of these signals demonstrated they were well correlated with respect to kinetics (Figures 7C,D) and potency (Figure 7E).

As described above, epitope mapping of the anti-FPN monoclonal antibodies revealed that antibodies either recognized linear sequences in the fifth EC loop of FPN, or they appeared to bind a conformational epitope (Table 1). The structure of FPN, based on the structure of the bacterial FPN homolog, BbFPN (Taniguchi et al., 2015), is proposed to comprise two transmembrane lobes and hepcidin is proposed to bind within the central cavity formed by these lobes (Figure 1). Because the fifth loop is not predicted to be involved in hepcidin binding, we did not anticipate that antibodies binding to the fifth EC loop would directly antagonize hepcidin binding; however, it was theoretically possible that the antibodies might stabilize FPN through a different mechanism. To determine if the monoclonal antibodies were capable of preserving FPN on the cell surface in the presence of hepcidin, we evaluated their activity using two different methods: iron response reporter (BLA) and RhoG-hepcidin uptake assays. Iron reporter cells were incubated with titrated anti-FPN antibodies 38C8 or 38G6, or control antibodies (IgG or anti-hepcidin; Sasu et al., 2010) for 1 h prior to addition of hepcidin at ~EC₇₀ (36 nM) for 18 h. Although the anti-FPN antibodies showed activity that was significantly different than control, they were not as effective as the anti-hepcidin positive control antibody at maintaining surface FPN expression. In this assay, the maximum activities of tested antibodies were ~700 nM for antibody 38G6, ~2 μM for antibody 38C8 and ~25 nM for the anti-hepcidin antibody (Figure 8A). The RhoG-hepcidin uptake assay is a more proximal assay than the iron response assay, but the optimal RhoG-hepcidin concentration required for detection (250 nM) is higher than the EC₇₀ value (36 nM) used in the iron response reporter assay. Cells were pre-incubated with antibodies for 1 h before addition of 250 nM RhoG-hepcidin. RhoG-hepcidin uptake was then measured 1.5 h later. Once again, partial protection was observed as Abs 38C8 and 38G6 reduced internalization by roughly 50% (Figure 8B). The protective effect was observed to be reduced with higher concentrations of hepcidin, although the concentrations used in both assays may exceed those observed in some diseased states (Ganz et al., 2008; Zaritsky et al., 2009; Macdougall et al., 2010).

In an effort to identify small molecule inhibitors of hepcidin-induced FPN internalization, we ran a small molecule screen using the cell-based iron response reporter assay described above. A library of ~300,000 small molecules was screened in a single-point assay and hits were titrated to determine their potency. Molecules that prevented intracellular iron accumulation were further tested in the RhoG-hepcidin uptake and FPN internalization assays. The uptake assay indicated that the small molecules had different modes of action; some compounds inhibited FPN internalization by blocking hepcidin binding and others inhibited FPN internalization without directly blocking hepcidin (Figure 9A). Compounds from the latter phenotypic group were found to be non-specific, as they were (1) active in a general endocytosis counterscreen and/or (2) analogs of the compounds were found to be inactive. One compound that inhibited RhoG-hepcidin uptake was a sulfonyl quinoxaline (Compound 1, IC₅₀ = 141 nM, Figure 9B). Activated heteroaromatic molecules like 2-sulfonyl quinoxalines can interact with nucleophiles (Sundaram et al., 2007); hence we evaluated the reactivity of compound 1 in the presence of biologically-relevant nucleophiles. Incubation of compound 1 and related 2-sulfonyl quinoxalines in our collection with glutathione showed formation of a glutathione adduct (data not shown). FPN contains a cysteine (C326) in the fourth loop that is reported to contain a free sulfur, i.e., not involved in a disulfide bond (Liu et al., 2005). Previous reports by Nemeth et al. (Fernandes et al., 2009) indicate that treatment of FPN with non-membrane permeable thiol-modifying reagents such as iodoacetamide prevents hepcidin binding, presumably through reaction with the free sulfur in C326. However, hepcidin also contains eight cysteines and while all of these cysteines are reported to be involved in disulfide bonds (Jordan et al., 2009), it is theoretically possible that the compounds could react with hepcidin if the disulfide bonds were labile. To determine whether the hepcidin-blocking effect of compound 1 was reversible, we tested compound 1 and iodoacetamide in a modified version of the RhoG-hepcidin uptake assay. In this instance we treated cells with compound 1 or iodoacetamide for 1 h and then washed the cells to remove any unbound compound; RhoG-hepcidin was then added and uptake assessed. In this format, the positive control iodoacetamide blocked RhoG-hepcidin binding with an IC₅₀ of 86 μM, while compound 1 was active at a significantly lower concentration with IC₅₀ value of 0.1 μM. Similar potencies were observed in experiments in which compound was continuously present (Figure 9C, left panel). These data indicate that the compounds bound irreversibly to target, in contrast to a control non-reactive, reversibly-binding compound, which was inactive after washout (Figure 9C, right panel). We confirmed the compounds bound irreversibly to FPN using a tritium-labeled analog of compound 1 (compound 2, Figure 9B). Compound 2 was incubated with uninduced and induced T-REx™Fpn-V5 cells and was only detectable in lysates from induced cells. Association of tritiated compound 2 with FPN was established by demonstrating that the compound 2-labeled band migrated at the same molecular weight as FPN, detected by Western analysis (Figure 9D).

The thiol reactivity of compound 1 was demonstrated using two linear peptides; one composed of the FPN sequence including C326 which is required for hepcidin binding (Figure 10A); the other was composed of the same sequence with the exception that C326 was replaced with the non-thiol amino acid tyrosine (Figure 10B). Adduct formation was assessed with mass spectrometry, using iodoacetamide as a control. Reactivity of compound 1 with the peptide was demonstrated by observation of adducts corresponding to the incorporation of the quinoxaline core via substitution of the sulfonyl moiety and was only observed when cysteine was present (Figure 10). We inferred that compound 1 potentially reacted with C326 in cell-expressed FPN.

To evaluate the selectivity of compound 1 for folded FPN vs. thiol reactivity, we compared the rate of reaction of compound 1 with cell-expressed FPN vs. the rate of reaction with the C326-containing linear peptide. A faster rate of reaction with cellular FPN vs. the linear peptide would suggest that compound 1 has some affinity for folded FPN and binding facilitates reaction with the thiol moiety of C326. Reactivity with cell-based, folded FPN was assessed by the hepcidin uptake assay (Figure 9C); reactivity with the linear peptide was assessed by mass spectrometry (Figure 10A). In cells, 100 μM compound 1 inhibited 100% of hepcidin uptake after just 1 h. In comparison, 100 μM compound 1 reacted with only 50% of the peptide in 12 h. This suggests that compound 1 has some affinity for folded FPN protein. Iodoacetamide, which has no affinity for folded FPN and was selected as a control because of its thiol reactivity, reacted completely with the cysteine-containing peptide within 1 h (100 μM iodoacetamide) whereas this same concentration only inhibited ~50% of hepcidin uptake within the same time frame.

It was previously reported that mammalian hepcidin did not bind to FPN at low temperatures (De Domenico et al., 2008). To verify this work, RhoG-hepcidin was incubated with either uninduced or induced cells at 0°C. Contrary to the previous report, RhoG-hepcidin bound to FPN at this low temperature. To assess the reversibility of hepcidin binding to FPN, we incubated FPN-expressing cells with RhoG-hepcidin at 0°C to prevent internalization; we then added excess unlabeled hepcidin to determine if the labeled hepcidin could be displaced. We found that excess unlabeled hepcidin effectively competed pre-bound RhoG-hepcidin from the surface of FPN-expressing cells under conditions where FPN/hepcidin could not be internalized, i.e., at 0°C. We confirmed that FPN was not internalized using antibody 45D8 (Figure 11A). These results demonstrated that RhoG-hepcidin bound FPN at 0°C and that the binding was reversible.

FPN has three potential N-linked glycosylation sites (N174, N434 and N567). One of these sites (N434) is in the fifth EC loop, the same loop to which our monoclonal antibodies bind. Tissue-specific differences in hepcidin activity in mice have been reported, and it was suggested that FPN glycosylation or other post-translational modifications may be responsible for these differences (Canonne-Hergaux et al., 2006; Drakesmith et al., 2015). To determine if glycosylation of human FPN is required for hepcidin-mediated internalization and degradation, we treated induced T-REx™/Fpn-V5 cells with tunicamycin, which blocks addition of N-linked glycosylation (Lehle and Tanner, 1976); cells were subsequently treated with hepcidin, which resulted in degradation of both glycosylated and non-glycosylated FPN, as shown by Western blot (Figure 11B). Similarly, we saw no difference in RhoG-hepcidin uptake with and without tunicamycin treatment (Figure 11C). We also discovered that one of our human monoclonal antibodies, antibody 38C8, did not recognize enzymatically deglycosylated FPN (Peptide-N-Glycosidase F (PNGase F)-treated). Western analysis revealed that antibody 38G6 recognized both glycosylated and deglycosylated FPN, but PNGase F-treated FPN was not detectable with antibody 38C8, even at the very high concentration of 20 μg/ml (Supplementary Figure 3A). When cells were treated with tunicamycin followed by hepcidin, we confirmed by immunoblot that (1) antibody 38G6, but not antibody 38C8, detected non-glycosylated FPN and (2) hepcidin-mediated FPN internalization did not require N-linked glycosylation of FPN (Supplementary Figure 3B). Detection of non-glycosylated FPN was also observed with antibody 38G6, but not antibody 38C8, following tunicamycin treatment in immunofluorescence experiments (Supplementary Figure 3C).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.