Footprinting mass spectrometry of membrane proteins: ferroportin reconstituted in saposin A picodiscs

Abstract

Membrane proteins participate in a broad range of cellular processes and represent more than 60% of drug targets. A new approach to their structural analyses is mass spectrometry (MS)-based footprinting including hydrogen/deuterium exchange (HDX), fast photochemical oxidation of proteins (FPOP), and residue-specific chemical modification. Studying membrane proteins usually requires their isolation from the native lipid environment, after which they often become unstable. To overcome this problem, we are pursuing a novel methodology of incorporating membrane proteins into saposin A picodiscs for MS footprinting. We apply different footprinting approaches to a model membrane protein, mouse ferroportin, in picodiscs and achieve high coverage that enables the analysis of the ferroportin structure. FPOP footprinting shows extensively labeling of the extramembrane regions of ferroportin and protection at its transmembrane regions, suggesting that the membrane folding of ferroportin is maintained throughout the labeling process. In contrast, an amphipathic reagent, N-ethylmaleimide (NEM), efficiently labels cysteine residues in both extramembrane and transmembrane regions, thereby affording complementary footprinting coverage. Finally, optimization of sample treatment gives a peptic-map of ferroportin in picodiscs with 92% sequence coverage, setting the stage for HDX. These results, taken together, show that picodiscs are a new platform broadly applicable to bottom-up mass spectrometry studies of membrane proteins.

INTRODUCTION

Membrane proteins participate in almost every physiological process in life. They represent approximately 25–30% of total proteins and are major targets for drug development. In vitro structural and functional studies of membrane proteins, however, have always been challenging. These proteins often denature, aggregate, and lose their biochemical activity in detergents used for protein extraction and purification. This is because detergent micelles often do not mimic a cell membrane that maintains the native conformation of membrane proteins. To overcome this problem, several lipid bilayer systems have been developed to stabilize membrane proteins in vitro. These include liposomes¹, nanodiscs², bicelles³, native lipid discs generated by styrene-maleic acid (SMA)⁴, and diisobutylene maleic acid (DIBMA) co-polymers⁵. Lipid-free systems, such as amphipathic polymers (amphipols), have also been widely used to maintain membrane proteins in aqueous solution⁶. Reconstitution of membrane proteins, however, especially unstable ones of eukaryotic origin, in these systems is often inefficient and requires extensive optimization, presenting an obstacle for the structural analyses of these proteins.

The saposin-lipoprotein nanoparticle system (Salipro), or picodiscs, is a novel and advantageous methodology for membrane protein reconstitution into lipid bilayers⁷. The saposin A peptides surrounding the lipid bilayer that forms picodiscs are soluble in aqueous solution (Figure 1A). To conduct the reconstitution, membrane proteins purified in detergent are mixed with phospholipids and saposin A. Subsequent dilution of this mixture below the critical micelle concentration (CMC) of the detergent allows the replacement of protein-bound detergent molecules by picodiscs, thereby promoting reconstitution into this lipid-bilayer system. The membrane protein in picodiscs can be subsequently purified through size-exclusion chromatography (SEC) in a detergent-free buffer. Unlike for liposomes or nanodiscs, assembling membrane proteins in picodiscs has a major advantage that detergent removal does not require solid-absorbent Bio-Beads whose use is a prolonged process that often causes membrane proteins to precipitate. Furthermore, the bead surface designed to absorb detergent also binds proteins nonspecifically, leading to significant reduction of the reconstitution efficiency. Another advantage of saposin A is that it exhibits high flexibility, adapting to the size of membrane proteins and forming picodiscs in various sizes⁷. In contrast, nanodiscs are limited by the size of membrane scaffold proteins surrounding the phospholipids. In sum, the picodiscs reconstitution allow efficient reconstitution of membrane proteins regardless of their sizes. Compared to nanodiscs, the reconstitution process of picodiscs is more straightforward and requires easier optimization of the reconstitution protocol.

Figure 1. Characterization of ferroportin-picodiscs.

A. The architecture of picodiscs with the lipid bilayer (green) surrounded by saposin (blue) and with embedded ferroportin (orange). B. The elution profile of purified ferroportin in DDM micelles (red) and in picodiscs (blue) on size-exclusion chromatography. C. SDS-PAGE with Coomassie staining. D. Negative-stain electron microscopy of ferroportin-picodiscs showing the top and side views.

Given these advantages of picodiscs, we reconstituted mouse ferroportin, a model membrane protein, into picodiscs for mass spectrometry footprinting analyses. Ferroportin belongs to the major facilitator superfamily (MFS) that represent one quarter of all transport proteins. Ferroportin is the sole iron exporter in vertebrates, and provides the central control of iron absorption, recycling, and storage. Crystallization of a eukaryotic variant of ferroportin has been difficult because its structure is flexible. Nevertheless, the crystal structures of a stable bacterial homolog of ferroportin was determined⁸. During the preparation of our manuscript, cryo-electron microscopy (cryo-EM) structures of human and primate ferroportin were reported^9,10. These EM structures, however, failed to show a few extramembrane regions of ferroportin. Nevertheless, these structures together afford a solid base for the interpretation of MS footprinting data.

Here we performed fast photochemical oxidation of proteins (FPOP)^11-15, and N-ethylmaleimide (NEM) specific amino acid footprinting^16,17 on ferroportin in picodiscs and used proteomics-based MS analyses to locate the labeled sites. We also performed footprinting on picodiscs and digested with pepsin to test the feasibility of applying the picodisc system for hydrogen-deuterium exchange (HDX)¹⁸, a well-established method for MS footprinting that has been used sparingly for membrane proteins^19-23. The footprinting of ferroportin-picodiscs with those common footprinting methods show high sequence coverage and extensive labeling. Furthermore, we generated a homology model of the full ferroportin structure by using MD simulation to validate the footprinting results. We find that incorporating membrane proteins into picodiscs is well suited for MS footprinting analyses.

RESULTS

Preparation of ferroportin in picodiscs.

The saposin A was expressed in E. coli and purified to near homogeneity. This scaffold protein is relatively unstable, and its storage requires care. In absence of lipids, saposin A at 1 mg/mL starting concentration slowly aggregates at 4 °C in buffers such as phosphate buffered saline. Within a few days, nearly all saposin A aggregates and loses its efficiency in binding lipids and forming picodiscs. The oligomerization occurs also after a freeze-and-thaw cycle for saposin stored at −80 °C. Therefore, in subsequent studies, we always use saposin freshly purified with size-exclusion chromatography for the picodisc assembly.

We expressed the mouse ferroportin in Pichia pastoris and purified the recombinant ferroportin in N-dodecyl-D-maltoside (DDM), a mild detergent commonly used for membrane protein purification. Milligrams of ferroportin could be obtained, and the protein is homogeneous on size-exclusion chromatography (SEC) and negative-stain electron microscopy. The mouse ferroportin, however, is highly unstable in detergent solution. We observed a large amount of protein precipitation after two days of purification. Such instability in detergents is commonly observed for membrane proteins, especially those of vertebrate origin. Reconstitution into a lipid bilayer system often can alleviate this problem.

Characterization of ferroportin-picodiscs and comparison with ferroportin-nanodiscs.

The freshly purified ferroportin was immediately reconstituted into saposin picodiscs (Figure 1A) and applied to SEC in a buffer without detergent. We found that the ferroportin reconstitution using bovine brain lipids or soybean lipids, compared to using synthetic lipids such as DOPC or POPC, shows a better SEC profile. The elution volume (13.2 mL on a Superdex 200 column) indicates the ferroportin-picodiscs are stable and not aggregated (Figure 1B). The peak appears symmetrical, suggesting the sample is relatively monodisperse and that most of the reconstitution product is comprised of ferroportin-picodiscs. Owing to the efficient reconstitution, the amount of empty picodiscs is small, as demonstrated by its elution volume at approximately 15 mL. The apparent size of ferroportin in picodiscs is like that of ferroportin in micelles, which emerges at an elution volume of 12.9 mL (Figure 1B). Because ferroportin is a monomeric protein, most picodiscs with ferroportin embedded should contain only one ferroportin molecule. On SDS-PAGE (Figure 1C), the ferroportin-picodisc complex shows protein bands representing ferroportin (~ 63 kDa) and saposin A (~ 10 kDa), as expected.

We performed negative-staining electron microscopy to visualize the picodiscs with ferroportin (Figure 1D). The ferroportin-picodiscs are found in single particles that adopte a variety of orientations, most being rod-shaped and disc-shaped, with average diameters of ~ 20 nm, respectively. These particle sizes are consistent with the size of ferroportin-picodiscs and the SEC elution volumes (Figure 1B). The rod-shaped particles correspond to a side view of picodiscs with ferroportin, and the disc-shaped particles correspond to a top view. The resolution of the negative-stain images, however, is insufficient to reveal structural details of the ferroportin-picodisc assembly. Taken together, SEC, SDS-PAGE and negative-stain EM evidence indicates that ferroportin-picodiscs are highly homogeneous.

Notably, the ferroportin-picodiscs remain stable and do not aggregate or precipitate after one month at 4 °C. As a control, we reconstituted mouse ferroportin into nanodiscs. The reconstitution is highly inefficient and very inconsistent in our hands. We observed a 90% loss of ferroportin protein during the lipid removal process. Furthermore, the ferroportin-nanodiscs precipitate within one week after reconstitution.

FPOP footprinting MS of ferroportin-picodiscs.

Compared to traditional structural biology methods, footprinting MS provides relatively low-resolution of structural information and often requires the combination of several complementary labeling methods to reach high coverage. Therefore, we included several alternative footprinting methods in our investigation of ferroportin in picodiscs.

We first labeled ferroportin-picodiscs with FPOP, a method that generates •OH radicals that actively react with the side chains of ~75% of amino-acids¹². After the irreversible labeling by FPOP, the ferroportin-picodiscs was applied to SDS-PAGE under denaturation conditions, which allow further enrichment of the footprinted ferroportin. Subsequently, we used a previously reported in-gel digestion protocol that is optimized for the high sequence coverage of membrane proteins¹⁷. The MS/MS spectra (Figure 2A) show that the number of ferroportin chymotryptic peptides affords 89-97% sequence coverage (Figure S1). Furthermore, the FPOP footprint is distributed over the entire protein except for a few short regions (Figure 2B). The FPOP modification sites are highly concentrated on the extramembrane regions, consistent with a previous report²⁴. In contrast, the transmembrane regions are labeled more sparsely, although most of these regions are covered by MS/MS of the peptides.

Figure 2. FPOP footprinting of ferroportin in picodiscs and DDM micelles.

A. Product-ion (MS/MS) spectrum of a representative peptide modified (red letter) after the FPOP footprinting. B. FPOP modification levels of peptides from ferroportin in picodiscs. C. Modification sites mapped onto the ferroportin homology model. D. FPOP modification levels of ferroportin in DDM micelles. E. Modification sites mapped onto the ferroportin homology model. The error bars are from experimental duplicates.

To assess further the conformational state of ferroportin in picodiscs, we compared the modification levels of all methionine residues. The methionines were selected because their δ-sulfur atom has high chemical reactivity with •OH radicals. Therefore, exposed methionines are expected to exhibit high oxidation extents, affording a signal-to-noise level better than those of other residues. Moreover, comparisons using this single residue ensure the nearly same intrinsic chemical reactivity and affords a relatively unbiased reporter for the exposure levels of methionine side chains located at different site and environments. For example, methionine buried inside the protein or protected by the lipid bilayer should show low modification levels. We found a highly contrastive pattern of methionine labeling in ferroportin. Several residues, including Met266, Met346, Met393, Met444, Met509, show the modification levels of over 20%. The highest modification was at Met266, at 52%. On the contrary, most other methionines are not modified, or modified at lower than 3%.

To understand these differences, we mapped onto a model of full-length mouse ferroportin (Figure 2C) based on the structure of a bacterial homolog⁸, which is consistent with the EM structure but shows most extramembrane regions. Remarkably, all the methionines showing low extents of modification are buried in the membrane (Figure 2B). In contrast, the highly modified methionines (>20%) are exposed in the extramembrane loops or at the membrane interface and have high solvent accessibility. For instance, the high modification level of Met346 indicates that this residue is solvent-accessible and reactive with •OH, consistent with its location at the exposed hepcidin-binding site. In addition, Trp82 (29%) and Trp299 (38%) are highly modified; these residues are located also in the extramembrane loops. Overall, these data are consistent with the proposition that ferroportin maintains its native folding topology in the membrane environment of picodiscs during reaction with the •OH radicals. Thus, picodiscs may provide a highly robust platform for MS footprinting.

To compare the outcome with picodiscs, we conducted FPOP labeling of ferroportin in detergent micelles and nanodiscs. Similar footprinting patterns are observed at the peptide level between picodiscs and freshly prepared samples in DDM micelles (Figure 2D, E). In our hands, however, FPOP of ferroportin reconstituted in nanodiscs did not give consistent footprinting results, despite repeated attempts. We attribute this result to significant loss of ferroportin protein during the sample preparation, and the low sample yield results in low sequence and footprinting coverages in subsequent MS analyses. Considering all the results, we conclude that the picodiscs system is a superior platform to nanodiscs and detergent micelles at least for stabilizing ferroportin, enabling high resolution and consistency for MS footprinting.

NEM footprinting MS of ferroportin-picodiscs.

Because FPOP preferably labels residues at the extramembrane regions, we used a membrane-permeable reagent, NEM, to provide cysteine footprints to complement those generated by FPOP. Furthermore, the use of NEM allows ready isotopic encoding (NEM/NEM-d₅) for the MS analysis to identify unequivocally the modifications through peak doublets and to quantify accurately the ratio between free and protected cysteines^17,25. We first incubated the ferroportin-picodiscs with NEM to label free cysteines that are not buried in the protein and do not form disulfide bonds. After generating the NEM footprints, the ferroportin protein was denatured and reduced to expose unmodified cysteines, which were subsequently labeled by NEM-d₅. We conducted duplicate analyses of the NEM labeling; these experiments (Figure 3A) show high consistency of NEM modification levels (Figure 3B), likely owing to the high reactivity of NEM to cysteines. MS analyses, aided by the isotopic encoding, identified multiple peptides that cover all the cysteines except Cys312. Most cysteines show high modification levels (>20%) and several are over 50%. For future applications, NEM efficiently labels Cys326, which is an essential residue for hepcidin to induce ferroportin internalization²⁶, allowing future experiments to detect the binding interactions of this important ligand.

Figure 3. Cysteine footprinting of ferroportin-picodiscs by using isotope-encoded NEM labeling.

A. MS/MS spectra of a representative peptide modified (red letter) with NEM/NEM-d5. B. NEM modification levels of peptides. C. The modification sites mapped onto the ferroportin homology model. The NEM modification percentage of peptides are calculated as NEM/(NEM-d₅ + NEM). Error bars are from experimental duplicates.

We mapped the NEM-labeled cysteines onto the homology model of ferroportin (Figure 3C). As expected, NEM efficiently labels cysteines in the transmembrane and soluble regions of ferroportin. Because the picodisc system is compatible with NEM labeling, we expect that other types of residue-specific irreversible labeling can be applied by using the picodisc platform. Like NEM footprinting, many of these residue-specific reagents have the advantage of using isotope labeling for accurate quantification²⁷. When used in combination, these reagents can target a spectrum of different residues.

Pepsin digestion of ferroportin-picodiscs for hydrogen–deuterium exchange analysis.

HDX MS footprinting requires rapid protease digestion and handling (within a few minutes) of protein samples to prevent back exchange. To evaluate the feasibility of performing HDX footprinting with picodiscs, we digested ferroportin-picodiscs with pepsin, a protease commonly used in HDX MS. To allow protein digestion in a short time window, we systematically optimized the digestion protocol under the HDX-quench conditions; optimized are the choices of protein denaturant, detergent to dissolve the picodiscs, method to remove phospholipids, online vs. offline pepsin digestion, fungal XIII to replace or augment pepsin, and the timing of each step. We found that the digestion efficiency is significantly improved with the use of RapiGest, a surfactant that facilitates lipid dissolution and protein unfolding. In addition, 4 M urea is optimal to unfold the ferroportin protein. We used ZrO₂ beads to remove phospholipids contained in the picodiscs, taking a lead from a previous report²⁸. We found there is no significant difference between using fungal XIII and pepsin or between online and offline pepsin digestion. All these steps, protein denaturation, lipid solubilization, lipid removal and protease digestion, can be accomplished within 3 min to meet the challenge of HDX MS. The combination of optimized unfolding and digestion steps significantly improves the mapping coverage. The MS analyses show that 92% of the entire ferroportin sequence was covered (Figure 4), setting the stage for future HDX experiments.

Figure 4. Sequence coverage of pepsin-digested ferroportin in picodiscs.

Black bars below the sequence indicate detected peptides.

DISCUSSION

Here we demonstrate that ferroportin, a membrane protein relatively unstable in detergent micelles, can be reconstituted in picodiscs for MS footprinting and then isolated and digested to give high sequence and footprinting coverage. To our knowledge, this is the first application of picodisc system for the MS footprinting of membrane proteins. Ferroportin appears to maintain its native folding topology during the footprinting process. We used two footprinting methods, FPOP and NEM, to illustrate that the picodisc system should be broadly applicable to bottom-up MS footprinting. The FPOP labeling occurs preferably at the extramembrane regions of ferroportin, whereas the NEM modification occurs also in intramembrane regions (Figure 5). In other words, the ferroportin-picodisc system is protective for FPOP footprinting of transmembrane regions of ferroportin, but remains accessible to NEM as a membrane-permeable reagent. Thus, these complementary labeling methods allow mapping with good coverage of the folding topology of ferroportin. The picodisc system should also work for MS footprinting with other labeling reagents or multiplex covalent-labeling cocktails.

Figure 5. Membrane topology of mouse ferroportin with modified sites.

FPOP labeled sites are shown in red spheres, unmodified methionines in green, and NEM labeled sites in brown.

Ferroportin is an example of many unstable membrane proteins that suffer from aggregation in detergent micelles and poor reconstitution with other lipid bilayer systems. In contrast, the picodisc system affords high efficiency of ferroportin reconstitution into the lipid bilayers. Under native-like lipid conditions, we find that this eukaryotic protein is highly stable, withstanding long storage times and remaining stable throughout the MS footprinting. The robust and feasible picodiscs system represents a valuable method for stabilizing unstable membrane proteins that quickly denature in detergent. The sample preparation of membrane proteins in picodiscs is rapid and straightforward.

We optimized a protocol for HDX-MS because rapidly unfolding and proteolysis of ferroportin in the tightly packed picodiscs is relatively difficult. The optimization protocol affords 92% sequence coverage of ferroportin within a short time window (3 min), which minimizes back exchange in HDX experiments. This high coverage show promise for future HDX studies of membrane proteins in picodiscs; such studies should report conformational changes, ligand interactions, and lipid modulations of membrane proteins.

Overall, the picodisc is an important and advantageous lipid bilayer system that enables proteomics studies of various membrane proteins. Our study illustrates the prospect of applying picodiscs for footprinting with reversible labeling (e.g., HDX), for broad-based modification of the protein backbone, and for irreversible labeling (FPOP, NEM) that modify the protein side chains. Together, these complementary approaches can be used in various scenarios ranging from studying the fast dynamics of membrane proteins to unraveling their native cellular states.

EXPERIMENTAL SECTION

Expression and purification of ferroportin.

Coding sequence of the full-length mouse ferroportin followed by a PreScission site, EGFP and 10X His tag, was cloned into a modified pPICZ-A vector for expression in Pichia pastoris. The Pichia cells were grown in BMG media media (1% glycerol, 0.34% yeast nitrogen base, 1% ammonium sulfate, 0.4 μg/mL biotin, and 100 mM potassium phosphate pH 6.0). The media was subsequently exchanged to BMM media without glycerol, and the cells were induced with 0.05% methanol for 48 h. The cells were collected through centrifugation and flash frozen in liquid nitrogen.

For protein purification, 20 g of frozen yeast cells were lysed with a PM100 ball mill (Retsch) to break cell walls. The lysed cells were resuspended in 60 mL buffer containing 50 mM Tris, pH 8.0, 150 mM NaCl, and protease inhibitors. To dissolve the cell membranes, 1.2 g DDM was added, and the suspension stirred at 4 °C for 3 h. The cell debris was removed by ultracentrifugation at 20,000 g for 45 min. The supernatant was incubated with 3 mL of Ni-NTA resin (Qiagen) at 4 °C for 3 h. The resin was subsequently washed with 50 mL buffer containing 25 mM imidazole, 150 mM NaCl, 0.05% DDM and 50 mM Tris, pH 8.0. The washed resin was incubated with 25 μg/mL PreScission protease at 4 °C overnight to remove the EGFP-10X His tag. The released ferroportin protein was concentrated and applied to size exclusion chromatography (Superdex200) that was equilibrated in a buffer containing 150 mM NaCl, 0.05% DDM and 50 mM Tris, pH 8.0. The peak fractions were concentrated to 10 mg/mL and used immediately for reconstitution into picodiscs.

Expression and purification of Saposin A.

The codon-optimized sequence of Saposin A with an N-terminal 6-His tag and a TEV cleavage site were cloned into a PET-28b plasmid. The plasmid was transformed into E. coli BL21 (DE3) cells. The cells were grown at 37 °C in LB media until 0.8 OD and induced with 1 mM IPTG for 3 h. The cells were collected by centrifugation, and pellets were flash frozen. Purification of Saposin A followed a published protocol ⁷.

Reconstitution of ferroportin into picodiscs.

Bovine brain lipids (Sigma-Aldrich) were dissolved in 1% DDM, 150 mM NaCl, and 50 mM Tris pH 8.0 and sonicated in a water bath at 25 °C for 5 min. The lipid solution was diluted to 5 mg/mL and mixed 1:1 (v/v) with 5 mg/mL purified ferroportin. The mixture was incubated at 37 °C for 5 min. Subsequently, purified saposin A was added to a final concentration of 0.5 mg/mL, and the mixture was incubated at 37 °C for 5 min. The mixture was diluted in 150 mM NaCl and 50 mM Tris pH 8.0 to lower the DDM concentration to 0.03% and favor the picodisc reconstitution. The mixture was incubated at room temperature for 5 min and loaded onto a Superdex 200 column in 150 mM NaCl and 50 mM Tris, pH 8.0.

Homology modeling and molecular dynamics simulation of ferroportin.

The homology model of ferroportin was generated by I-TASSER ²⁹ by using the crystal structure of a bacterial homolog of ferroportin in the outward-facing conformation (PDB code: 5AYN) as the template. The transmembrane region of this homology model superimposes well with the EM structures of ferroportin, and the full model contains extramembrane regions not observed in the EM structures. The model was inserted into a POPC bilayer by using CHARMM-GUI ^30,31. Molecular dynamics simulation of this system was performed with GROMACS MD simulation package (verion 5.0.4)³² with CHARMM36m force field ³³.

Negative-staining electron microscopy of ferroportin-picodiscs.

Ferroportin pidodiscs samples were diluted to 50–80 nM and stained in a 2% (w/v) solution of uranyl acetate following the standard deep-staining procedure on glow-discharged holey carbon-coated EM copper grids. The negatively stained specimens were visualized by a JEOL JEO-1400 Plus transmission electron microscope operated at 120 kV. Magnified digital micrographs of the specimen were taken at a nominal magnification of 40,000 on an AMT CCD camera with a pixel size of 2.29 Å at the specimen level. The defocus values used were −1.0-2.5 μm, and the total accumulated dose at the specimen was approximately 70 electrons per Ų.

FPOP labeling and in-gel digestion.

Ferroportin (> 2 μM) reconstituted in picodiscs was mixed with 20 mM glutamine (scavenger) and 20 mM H₂O₂ immediately before FPOP ³⁴. To generate hydroxyl radicals, H₂O₂ photolysis was initiated with a KrF excimer laser (GAM Laser Inc) at 248 nm. The sample flow rate was adjusted to give 20% irradiation-excluded volume and to minimize repeated laser exposure. The total time for one sample to pass through the FPOP system was 120 s. After the laser irradiation, the collected samples were mixed with 10 mM catalase and 20 mM methionine (final concentrations) to decompose residual H₂O₂. The labeled samples were immediately loaded on SDS-PAGE. In-gel digestion followed a previously reported protocol that was optimized for high sequence coverage of membrane proteins ¹⁷. Briefly, smashed gel pieces were reduced with 5 mM Tris(2-carboxyethyl)phosphine (TCEP), and free thiols were blocked by reaction with 20 mM iodoacetamide. Chymotrypsin digestion was performed in presence of RapiGest™ (Waters).

NEM labeling and in-gel digestion.

The ferroportin-picodiscs (> 2 μM) was treated with 10 mM NEM (final concentration) for 15 min on ice. The labeled samples were immediately loaded onto SDS-PAGE. During in-gel digestion, 10 mM deuterated NEM (NEM-d₅) was used to block remaining free cysteines.

LC-MS/MS analyses.

The digested ferroportin peptides were injected into a C18 reversed-phase trapping column (Acclaim® PepMap100, 100 μm/2 cm, 5 μm, 100 Å; Thermo) that was running at 4.5 μL/min with 0.1% formic acid in water. The peptides were separated using a custom-packed C18 reversed-phase column (Magic, 100 μm/160 mm, 5 μm, 200 Å; Michrome Bioresource Inc) in a 125 min gradient of 0-80% acetonitrile at 500 nL/min. The scan range of precursor ions was set from 380 to 2200 (m/z) for all samples at high mass resolving power of 70000 at m/z 200. The mass spectrometer was operated in non-targeted data-dependent acquisition (DDA) mode. Eighteen most abundant precursor ions eluting per scan were subjected to high-energy C-trap dissociation (HCD) to give product-ion (MS/MS) spectra. Dynamic exclusion was activated for 8 s following each scan to enable selection and fragmentation of low abundance peptides. The LC-MS/MS analyses were conducted with the UltiMate 3000 Nano LC system and a Thermo Q Exactive Plus Hybrid Quadruple-Orbitrap mass spectrometer (Thermo).

Analyses of MS footprinting data.

The raw MS data were searched by using Byonic (Protein Metrics Inc). All known HO• side-chain reaction products and IAA alkylation were added to the variable modification database for the mass search of HO•-modified peptides. The search tolerance window was 15 ppm for precursor ions, and 50 ppm for product ions. Aromatic amino acids F, W, Y, L, M, H were selected as the cleavage sites for chymotryptic proteolysis. Quantitation of modification level was based on ion abundances from extracted ion chromatograms (XICs). The fractions modified at the peptide and residue levels were calculated as follows: time-dependent signals from LC-MS for each unmodified peptide and its modified species were extracted from raw data files by using XcaliburTM Software (Thermo) with a mass tolerance of 15 ppm. The peak integrals of both unmodified species and modified species were readily attainable. The modified fraction of each peptide was determined by dividing the total area of modified species by the sum of areas of the modified and unmodified species.

HDX and Pepsin digestion of ferroportin-picodiscs.

Ferroportin-picodiscs (>2 μM) were mixed 2:3 (v/v) with quench buffer containing 4 M urea, 0.5 or 1% trifluoroacetic acid, pH 2.5 on ice. Pepsin (0.33 mg/mL final concentration) and RapiGest (0.1 mg/mL) were added immediately, and the digestion was carried out on ice for 3 min. Subsequently, 3 mg of ZrO₂ beads (Supelco/Sigma) were added. After 2 s vortex, the sample was filtered by a disposable acetyl bio-cellulose filter for 30 s, and the flow-through was immediately injected into a custom-built HDX flow system. The samples were passed through an immobilized pepsin column (2 mm × 20 mm) at 100 μL/min flow rate in 0.1% formic acid. The digested peptides were passed through a ZORBAX Eclipse XDB C8 column (2.1 mm × 15 mm, Agilent) for 3 min desalting, and were subseqently directed onto an Hypersil Gold C18 column (2.1 mm × 50 mm, Thermo) with a 15 min gradient of 15%-80% acetonitrile in 0.1% formic acid at 50 μL/min flow rate.

The separated peptides were directed to an LTQ Orbitrap XL mass spectrometer (Thermo Fisher) equipped with an electrospray ionization source, and data and data analysis was performed using HDExaminer (Sierra Analytics). Prior to the HDX, the LTQ Orbitrap XL mass spectrometer was used in data-dependent fragmentation mode, monitoring the six most abundant peptides, to establish a map for the HDX data acquisition. Data were analyzed for sequencing and accurate precursor mass (± 5 ppm) using Byonic (Protein Metrics), and the peptides were manually curated to guide the HDX study.