Bile from the hemojuvelin-deficient mouse model of iron excess is enriched in iron and ferritin

Abstract

Iron is an essential nutrient but is toxic in excess. Iron deficiency is the most prevalent nutritional deficiency and typically linked to inadequate intake. Iron excess is also common and usually due to genetic defects that perturb expression of hepcidin, a hormone that inhibits dietary iron absorption. Our understanding of iron absorption far exceeds that of iron excretion, which is believed to contribute minimally to iron homeostasis. Prior to the discovery of hepcidin, multiple studies showed that excess iron undergoes biliary excretion. We recently reported that wild-type mice raised on an iron-rich diet have increased bile levels of iron and ferritin, a multi-subunit iron storage protein. Given that genetic defects leading to excessive iron absorption are much more common causes of iron excess than dietary loading, we set out to determine if an inherited form of iron excess known as hereditary hemochromatosis also results in bile iron loading. We employed mice deficient in hemojuvelin, a protein essential for hepcidin expression. Mutant mice developed bile iron and ferritin excess. While lysosomal exocytosis has been implicated in ferritin export into bile, knockdown of Tfeb, a regulator of lysosomal biogenesis and function, did not impact bile iron or ferritin levels. Bile proteomes differed between female and male mice for wild-type and hemojuvelin-deficient mice, suggesting sex and iron excess impact bile protein content. Overall, our findings support the notion that excess iron undergoes biliary excretion in genetically determined iron excess.

Graphical Abstract

Graphical Abstract.

Under conditions of iron excess, hepatocytes secrete iron-rich ferritin into bile and iron-poor ferritin into blood.

Introduction

Iron is an essential dietary nutrient.^1–3 Most iron in the body is dedicated to oxygen transport by hemoglobin. Given its capacity for oxidation/reduction, iron is toxic in excess. Toxicity manifests as cirrhosis, liver cancer, heart disease, arthritis, and diabetes. Iron-related conditions are common worldwide. Iron deficiency is the most prevalent nutritional deficiency and typically linked to inadequate intake. Iron excess is also common and usually due to genetic defects that directly or indirectly perturb expression of hepcidin, a hormone that inhibits dietary iron absorption. Produced largely by the liver, hepcidin post-translationally inhibits ferroportin, a transporter that exports iron from enterocytes and macrophages into blood. Hepcidin deficiency leads to unabated ferroportin-mediated iron absorption. Examples of diseases of iron excess include hereditary hemochromatosis and β-thalassemia. Iron excess can also result from blood transfusions, a common treatment for sickle cell disease and other anemias. Iron excess is treated with phlebotomy, unless contraindicated by anemia and chelators, although the latter can have adverse effects such as gastrointestinal distress and cytopenia.³ Hepcidin analogs and mimetics are under evaluation as treatments for iron excess.^4,5

Iron absorption is a central regulator of iron levels. Multiple molecular determinants of iron absorption, including ferroportin and hepcidin are known.^1–3 In contrast, iron excretion is poorly understood. It is believed to contribute minimally to iron homeostasis, even though a 1968 study observed that a human population with greater iron stores excreted more iron.⁶ Excretion is attributed to exfoliation of dead skin, turnover of intestinal epithelium, and blood loss from menstruation or minor epithelial trauma. However, prior to the discovery of hepcidin, studies showed that iron undergoes biliary excretion under conditions of iron excess, although not all studies found this.^7–20 Other studies predating the discovery of hepcidin also demonstrated that iron can undergo enterohepatic circulation, the process by which substances excreted into bile are reabsorbed by the intestines for return to the liver.^7,9,10 The relevance of biliary excretion and enterohepatic circulation to iron homeostasis has yet to be fully interrogated.

We recently reported that wild-type mice raised on an iron-rich diet have increased bile iron levels and that uptake of excess iron into hepatocytes by the metal transporter Zip14 (Slc39a14) is an essential prerequisite for subsequent excretion of iron into bile.²¹ In bile from iron-loaded wild-type mice, iron was present in two forms, heme and ferritin. Heme, also known as iron protoporphyrin IX, is an essential component of hemoglobin and other proteins required for gas transport and sensing, oxidative metabolism, and xenobiotic detoxification.²² Ferritin, an iron storage protein, is composed of multiple heavy and light chain subunits and sequesters iron in a redox-inactive form.²³ Ferritin expression is increased with iron excess and inflammation and decreased with iron deficiency. In conditions of iron excess, ferritin is not only enriched in serum, albeit in an iron-poor form, but also in bile although not all studies support this.^{11–16,18–20,24}

Our report demonstrated that dietary iron excess leads to increased bile iron levels, supporting the notion that iron undergoes biliary iron excretion under conditions of iron excess. However, genetic defects leading to excessive dietary iron absorption, such as those seen in hereditary hemochromatosis, are much more common causes of iron excess than dietary iron loading. In this study, we set out to determine if hereditary hemochromatosis also results in bile iron loading. To this end, we employed mice with deficiency in hemojuvelin (Hjv), a membrane protein essential for hepcidin expression by the liver.²⁵ These mice develop hepcidin deficiency, excessive dietary iron absorption, and systemic iron excess. Bile was collected surgically from these mice and analyzed for iron and ferritin levels, ferritin glycosylation, and protein abundance using proteomics. Hjv^−/− mice were also subjected to AAV-mediated knockdown of Tfeb, the master regulator of lysosomal biogenesis, function, and exocytosis, and analyzed for bile iron and ferritin content, given that lysosomal exocytosis has been implicated in bile ferritin excretion.

Methods

Generation and treatment of mice and sample collection

Animal work was approved by the Institutional Animal Care and Use Committee at Brown University. Hjv^± mice on the C57BL/6 N background were obtained from Jackson Laboratory (#017 788). Heterozygous mice were bred together to generate wild-type (Hjv^+/+) and mutant (Hjv^−/−) mice. Mice were housed in ventilated cage racks, maintained on a 12-h light/dark cycle with controlled temperature and humidity, and provided LabDiet 5010 chow and water ad libitum. Littermates were randomly assigned to experimental groups. Adeno-associated viruses (AAV8 serotype) were purchased from Vector Biolabs and injected retro-orbitally into mice under anesthesia in a single dose of 2 × 10¹¹ genome copies per mouse.

To collect samples from mice, mice underwent bile, blood, and tissue collection as previously described, with bile collected by ligation of the common bile duct, cannulation of the gallbladder, and collection over 60 min.²⁶ Blood was collected via inferior vena cava puncture into serum collection tubes (BD). Mice were euthanized by cervical dislocation followed by tissue collection for biochemical analysis. Tissue samples were flash-frozen in liquid nitrogen and stored at −80°C until analysis.

Non-heme iron measurements

Measurements were performed using the bathophenanthroline-based colorimetric assay.²⁷ Tissues (10–200 mg) were digested in 1 ml 3 N hydrochloric acid (Fisher)/10% trichloroacetic acid (Millipore Sigma) at 65°C for 2 days, with 30 min vortexing each day, followed by centrifugation. Iron levels were measured by mixing 10 μl supernatants with 200 μl chromagen in a 96-well plate. Chromagen consisted of five volumes MilliQ water, five volumes saturated sodium acetate (Fisher), and one volume chromagen stock, which comprised 0.1% bathophenanthroline sulfonate (Millipore Sigma) and 1% thioglycolic acid (Millipore Sigma). Iron standards (Fisher) were included. After a 10-min incubation, absorbances were measured at 535 nm. Mock digests without samples were included for this and all other metal analyses.

Protein analysis (non-proteomic)

To analyze proteins using gel electrophoresis, 50–200 mg of liver from each mouse were homogenized in 50 mM Tris pH 7.4/150 mM NaCl/1% Triton X-100/5 mM EDTA/Halt protease inhibitors (Thermo Fisher Scientific), then centrifuged. Supernatants were assayed for protein using DC Protein Assay (BioRad). For iron-stained gels, samples were electrophoresed on Criterion 4–20% TGX Stain-Free Gels (BioRad) under native conditions and imaged for protein using a ChemiDoc (BioRad). Gels were washed with MilliQ water for 5 min four times, stained for 10 min using Iron Stain Kit (MilliporeSigma), imaged using a ChemiDoc, washed, stained using Sigmafast DAB with Metal Enhancer (MilliporeSigma) for 30 min, washed, then imaged. For denaturing immunoblots, samples were electrophoresed on Criterion 4–20% TGX Stain-Free Gels under denaturing, reducing conditions, and imaged for protein. Gels were transferred to PVDF membranes (Thermo Fisher Scientific) overnight using Criterion Blotter (BioRad), then blocked for 1 h in 5% nonfat skim milk/TBST. Blots were incubated with Proteintech anti-Ftl1 #10727-1-AP or Cell Signaling anti-Fth1 #3998S at 1:1000, Proteintech anti-GAPDH #10494-1-AP at 1:5000, Proteintech anti-transferrin #17435-1-AP at 1:1000, Alpha Diagnostics anti-transferrin receptor 2 #TFR21-A at 1:1000, or with Fortis anti-TFEB antibody #A303-673A at 1:1000 for 2 h, washed with TBST for 5 min four times, incubated with Proteintech anti-rabbit HRP-linked IgG #SA00001-2 at 1:5000 for 1 h, washed, then imaged using Amersham ECL Prime (Cytiva) and a ChemiDoc.

To measure Flt1, hemoglobin, and heme levels, bile samples were analyzed according to manufacturer's instructions using Mouse Ferritin ELISA Kit for Ftl1 (Abcam), Mouse Hemoglobin ELISA Kit (Abcam), and Heme Assay Kit (Abcam). To assess protein glycosylation, samples were treated with PNGase F (NEB) according to manufacturer's instructions and then analyzed by immunoblot.

Proteomics

An amount of 30 μg of protein from each bile acid sample was removed and combined with approximately 2 volumes of lysis buffer, consisting of 5% sodium dodecyl sulfate (Gibco) in 50 mM triethylammonium bicarbonate buffer pH 7.5 (TEAB, Sigma–Aldrich). Samples were subjected to S-Trap Micro Digestion and Clean-up (ProtiFi, C02-micro-40) as specified by the manufacturer with some modifications. Samples were reduced using 10 mM dithiothreitol (Fisher Scientific) and incubated at 55°C for 45 minutes with mixing. Following incubation, samples were alkylated using 20 mM iodoacetamide (Thermo Fisher Scientific) and incubated at room temperature in the dark for 30 min. Following alkylation, samples were acidified using 2.5% phosphoric acid (Sigma–Aldrich) and vortexed for 30 s. Six volumes of ice-cold binding buffer, consisting of 90% LC/MS-grade methanol (ACROS Organics) and 10% 1 M TEAB, was added to each sample before the entire sample was transferred to an S-Trap spin column and centrifuged for 30 s at 4000×g. Each sample was then washed three times with 150 μl binding buffer. Following washing, trypsin (Promega Corporation) was reconstituted in 50 mM TEAB and added to each micro spin column in a 1:50 enzyme to protein ratio. Samples were then moved to a humidified chamber in a 37°C incubator and left overnight. On day 2, samples were left to cool to room temperature for 15 min before 40 μl of 50 mM TEAB was added and centrifuged for 1 min at 4000×g. Peptides were eluted with 40 μl LC/MS grade water (Honeywell) and 0.1% formic acid (Fisher) and centrifuged again for 1 min at 4000×g. Finally, to completely elute hydrophobic peptides, 40 μl 50% LC/MS grade acetonitrile (Thermo Fisher Scientific) was added to each sample and allowed to stand for 5 min before being centrifuged for 1 min at 4000×g. Each sample was concentrated using a speed vacuum concentrator (Thermo Fisher Scientific) for approximately 90 min. With solvent removed, samples were reconstituted in 100 μl solvent A, consisting of water with 0.1% formic acid, and spiked with indexed retention time peptides (Biognosys) in a 1:50 dilution to monitor HPLC performance.

To perform liquid chromatography-mass spectrometry (LC-MS) proteomic analysis, samples were then injected onto a QExactive Orbitrap LC/MS system (Thermo Fisher Scientific) and separated over a 120-min gradient consisting of solvent A and solvent B, consisting of acetonitrile and 0.1% formic acid, using a split-flow set-up on an Agilent 1200 series HPLC system. A 15 cm long, 75 μm inner diameter capillary analytical column packed with XSelect CSH C18 2.5 μm resin (Waters) from the IDeA National Resource for Quantitative Proteomics at the University of Arkansas for Medical Sciences (UAMS) was used to separate digested peptides. The 120-min gradient consisted of 95% solvent A for 1 min, 70% solvent A for 94 min, 5% solvent A for 6 min, and 100% solvent A for the remaining 20 min all at a flow rate of 0.190 ml/min. The mass spectrometer acquisition utilized a Full MS/ddMS2 (Top9) centroid experiment. Full MS parameters used a default charge state of 2, 1 microscan, resolution of 70 000, AGC target of 3e6, 200 ms maximum injection time, in a scan range of 400–1800 m/z. ddMS2 parameters were 1 microscan, resolution of 17 500, AGC target of 2e4, a max injection time of 200 ms, a loop count of 9, top nine precursors, isolation window of 2.5 m/z, collision energy of 28, and a scan range of 200–2000 m/z. Data-dependent settings consisted of a minimum AGC target of 2e2, 1e3 intensity threshold, unassigned charge exclusion, all charge states, preferred peptide match, isotope exclusion, 30 s of dynamic exclusion.

To perform label-free quantitation and MaxQuant searches, .raw files were processed using MaxQuant Ver 2.1.4.0 using 1% peptide and protein FDR search constraints. All values were default except for 20 ppm first search tolerance and 7 ppm main search tolerance against the latest version of the UniProt protein .fasta file for Mus musculus, canonical with isoforms, 20230316. Variable modifications consisted of deamidation (N), oxidation (M), and protein N-terminal acetylation. Carbamidomethylation was the sole static modification for the Trypsin/P search. Match between runs algorithm was unselected and intensity-based absolute quantification (iBAQ) was selected. The proteingroups.txt file output was further processed in MS Excel and relative IBAQ was calculated as the intensity of each protein divided by the sum of all protein intensities for each given sample.^28,29

Ontology was performed using ShinyGO at http://bioinformatics.sdstate.edu/go/.³⁰ Venn diagrams were generated using bioinformatics.psb.ugent.be/webtools/Venn/. Heatmaps were generated using Morpheus at https://software.broadinstitute.org/morpheus/. Uniprot (https://www.uniprot.org/) was used to assign bile proteins to specific biochemical pathways.

Statistical analysis

Statistics for all studies except proteomics were performed using GraphPad Prism 9 v9.3.1. Data were first tested for normal distribution by Shapiro–Wilk test; If not normally distributed, data were log transformed. Groups were compared by unpaired, two-tailed t tests. Data are shown as means ± standard deviation in the figures.

Results

We previously demonstrated that mice raised on an iron-rich diet have increased levels of iron and ferritin in bile.²¹ However, dietary iron loading is not a common cause of iron excess. To determine if iron and ferritin are enriched in bile in a more biomedically relevant context than dietary iron loading, we employed Hjv^−/− mice, a mouse model of hereditary hemochromatosis, a common inherited cause of iron excess due to excessive dietary iron absorption. Bile was collected from anesthetized mice by ligation of the common bile duct, cannulation of the gallbladder, and collection for 1 h. Mice were then subjected to blood collection and euthanized prior to tissue collection. As expected, liver non-heme iron levels were increased in female and male Hjv^−/− mice relative to Hjv^+/+ mice (Fig. 1A). Non-heme iron levels were also increased in bile from mutant mice (Fig. 1B). Ratios of bile to liver non-heme iron levels did not differ between Hjv^+/+ and Hjv^−/− mice, indicating that bile iron excretion is proportional to liver iron loading in this model (Fig. 1C). Bile ferritin light chain (Ftl1) levels were also increased in Hjv^−/− mice (Fig. 1D). Hemoglobin levels did not differ between Hjv^+/+ and Hjv^−/− mice, suggesting that increased biliary iron levels in mutant mice did not reflect contamination of collected bile with blood (Fig. 1E). Heme levels did not differ between female Hjv^+/+ and Hjv^−/− mice and were mildly increased in male Hjv^−/− mice compared to Hjv^+/+ mice, suggesting that increased heme iron export did not contribute prominently to bile iron excess in mutant mice (Fig. 1F). There was insufficient bile to analyze bile Fth1 and hemin levels after completion of other analyses described below.

Fig. 1.

Hjv^−/− mice have increased bile iron (Fe) and ferritin levels. Two-month-old Hjv^+/+ and Hjv^−/− were analyzed for: liver (A) and bile (B) non-heme Fe levels; ratios of bile to liver non-heme Fe levels (C); bile Flt1 levels (D); bile hemoglobin levels (E); bile heme levels (F). For all panels: at least four samples per group (except for female Hjv^+/+ samples in (F) due to limited sample volume); bars indicate mean ± standard deviation; genotypes compared by unpaired t-test (^*P<0.05).

To determine if bile ferritin levels were sufficiently elevated in Hjv^−/− mice to account for all non-heme Fe, we estimated the amount of iron that could be bound to ferritin in Hjv^−/− bile. In female Hjv^−/− mice, bile non-heme iron and Ftl1 levels were ∼50 and ∼275 μg/ml respectively, which equated to ∼895 and ∼13 μM assuming a molecular weight of 21 kD for Ftl1. If bile ferritin consists of 24 subunits and all non-heme Fe was ferritin-bound, each bile holo-ferritin would need to bind ∼1500 iron atoms. Ferritins can bind up to 4000 atoms of iron. This calculation suggested that bile ferritin levels were sufficient to bind all non-heme iron present in Hjv^−/− bile. A similar analysis for male Hjv^−/− mice yielded the same conclusion.

To compare the amount of heme and non-heme iron present in bile, we estimated the amount of iron present in heme in bile. In female Hjv^−/− mice, heme levels and non-heme iron levels were ∼81 and ∼49 μM respectively. In male Hjv^−/− mice, heme levels and non-heme iron levels were ∼68 and ∼36 μM, respectively. This indicated that heme iron is a prominent component of biliary iron in Hjv^−/− mice, similar to what we observed previously in wild-type mice raised on iron-rich diets.²¹

We next analyzed ferritin levels in liver, serum, and bile by gel electrophoresis. Holo-ferritin levels were analyzed using our previously validated protocol, which involves native gel electrophoresis followed by potassium ferrocyanide treatment with and without diaminobenzidine enhancement.²¹ Ftl1 and Fth1 levels were analyzed by immunoblots. As expected, Hjv^−/− liver lysates had prominent levels of holo-ferritin, Ftl1, and Fth1 (Fig. 2A). Hjv^−/− sera had undetectable levels of holo-ferritin and Fth1 in all samples and a mild increase in Ftl1 levels in some samples (Fig. 2B). These results were consistent with the long-standing observation that iron-poor ferritin, consistently largely of light chain subunits, is enriched in serum in conditions of iron excess. In contrast to sera, Hjv^−/− bile had increased holo-ferritin, Ftl1, and Fth1 levels although levels did vary between individual mice, particularly for male mutants (Fig. 2C).

Fig. 2.

Hjv^−/− mice have increased bile ferritin levels. (A–C) Liver (A), serum (B), and bile (C) from 2-month-old Hjv^+/+ and Hjv^−/− were analyzed by: native PAGE of 100 μg liver lysate, 10 μl bile, or 1 μl serum and Fe staining without (−DAB) and with (+DAB) enhancement (top two panels); denaturing, reducing immunoblot of 25 μg liver lysate or 1 μl bile or serum with anti-Ftl1 or Fth1 antibodies (bottom two panels). (D) liver lysates (5–25 μg), bile (1 μl), and serum (0.01–1 μl) from 2-month-old Hjv^−/− mice were treated with or without PNGase F, then analyzed by denaturing, reducing immunoblot for transferrin, transferrin receptor 2, Ftl1, and Fth1. For livers, 25 μg was analyzed for transferrin and transferrin receptor 2 blots; 5 μg was analyzed for Ftl1 and Fth1 blots. For serum, 1 μl was analyzed for Ftl1, Fth1, and transferrin receptor 2 blots; 1 μl of 1:100 dilution was analyzed for transferrin blot.

Two Ftl1 and Fth1 bands were noted in immunoblots of bile, compared to one band for both proteins in Hjv^−/− liver. This suggests that bile ferritin is biochemically distinct from liver ferritin in this mouse model. In our study of bile ferritin in wild-type mice on iron-rich diets, we also observed that bile and liver Ftl1 and Fth1 proteins migrated differently in denaturing, reducing conditions. We considered that differences in glycosylation may contribute to differences in migration of Ftl1 and Fth1. Hjv^−/−liver lysates, bile, and sera from mice were treated with and without PNGase F then analyzed by denaturing, reducing immunoblots. PNGase F treatment increased mobility of transferrin in liver, bile, and sera and transferrin receptor 2 in liver (Fig. 2D). In contrast, PNGase F treatment had no impact on migration of Ftl1 or Fth1 in liver or bile, indicating that differences in glycosylation do not contribute to size differences for ferritin subunits between bile and liver.

To further explore the impact of inherited iron excess on bile, we performed LC-MS on trypsin-digested samples from Hjv^+/+ and Hjv^−/− mice. We first compared female Hjv^−/− and Hjv^+/+ bile. Proteomic analysis detected 2250 proteins, with 247 total proteins differentially expressed between genotypes (Fig. 3A). Of these, 40 were more abundant and 207 less abundant in Hjv^−/− mice relative to Hjv^+/+ mice. Ftl1 and Fth1 were increased in female Hjv^−/− bile, consistent with immunoblot findings presented above. We did not detect differences in levels of hemoglobin subunits, transferrin, or lactoferrin between Hjv^+/+ and Hjv^−/− mice. Two other proteins of interest were decreased in female mutant bile. Aco1, also known as iron regulatory protein 1 (Irp1), encodes a dual-function protein that possesses aconitase activity when it binds an iron–sulfur cluster but attains RNA-binding activity when devoid of the cluster and impacts expression of genes governing cellular iron levels.³¹ Alad, delta-aminolevulinic acid dehydratase, is an iron–sulfur cluster-dependent protein that catalyzes the second step in heme biosynthesis.²² Neither Aco1 nor Alad is known traditionally as bile proteins.

Fig. 3.

Female Hjv^−/− and Hjv^+/+ mouse bile differ in protein content. LC-MS analysis was performed on 30 μg of bile from female Hjv^+/+ and Hjv^−/− mice (n = 5/group). (A) Volcano plot of female Hjv^−/− vs. Hjv^+/+ mouse bile. Differentially abundant proteins (adjusted P value < 0.05 and absolute value of log2(fold change)>1) are shown above −log10P=1.3 with protein names shown adjacent as space permitted. x–y coordinates of additional proteins with relevance to Fe homeostasis are shown in box; top two proteins are differentially expressed. Genes with log2(fold change)>0 are more abundant in Hjv^−/− mice. (B) Ontology of proteins more abundant in female Hjv^−/− bile. (C and D) Heatmap (C) and description (D) of lysosome-related proteins more abundant in female Hjv^−/− bile. (E and F) Ontology (E) and network analysis (F) of proteins less abundant in female Hjv^−/− bile. (G) Heatmap of proteins aligning with metabolic pathways in female Hjv mice.

In our published analysis of wild-type mice raised on iron-rich diets, we employed tandem mass tagging and LC-MS/MS to compare bile proteomes of female mice raised on iron-rich and -sufficient diets. We compared our proteomic results from that study with results from this study. Of the 38 proteins upregulated in bile from mice on the iron-rich diet, 7 were upregulated (Smpdl3a, Rgn, Lipa, Sod1, Enpep, Gns, and Ftl1), and 4 were downregulated (Hsd17b4, Adh1, Scp2, and Ctrc) in bile from Hjv^−/− mice. Of the 15 proteins downregulated in bile from mice on the iron-rich diet, one was downregulated (Uqcrc1) and none were upregulated in bile from Hjv^−/− mice. Overall, although the tandem mass tagging approach yielded fewer hits than this current study, there was not much overlap in proteomic datasets, suggesting that dietary and genetic iron loading do not have similar impacts on bile proteomes.

We next considered the 40 proteins more abundant in female Hjv^−/− bile. Pathway analysis indicated that 12 of 40 or 30% were lysosome related (Fig. 3B–D). While the molecular basis of biliary ferritin secretion is not established, older biochemical and ultrastructural studies implicated lysosomal exocytosis from hepatocytes.^{11–16,24,32–34} Notably, the transcription factor EB (TFEB), a key regulator of lysosomal biogenesis, autophagy, and exocytosis, has been associated recently with iron homeostasis, although a clear picture has yet to emerge.^35,36 TFEB increases lysosomal exocytosis of iron and attenuates cytotoxicity in iron-treated HEK293T cells³⁷ and stimulates hepcidin expression in hepatocyte cell lines.³⁸ Iron treatment increases TFEB expression in some cell lines but decreases it in others.^37,39,40 TFEB overexpression increases ferritin expression in cell lines and in mice.⁴¹ However, a role for TFEB in biliary iron or ferritin excretion has yet to be investigated but is explored below. Lysosomes also play a key role in ferritinophagy, the process by which ferritin is trafficked to the lysosome for degradation and release of iron stores into the cell.⁴² Since this process results in ferritin degradation, we speculate that ferritinophagy does not play a prominent role in transport of intact holo-ferritin into bile.

While 40 bile proteins were more abundant in female Hjv^−/− than Hjv^+/+ mice, 207 proteins were less abundant in mutant mice (Fig. 3A). Of the 207 proteins, 80 (39%), aligned with metabolic pathways (Fig. 3E, F). These 80 proteins associated largely with lipid biosynthesis, electron transport chain, and carbohydrate and amino acid metabolism (Fig. 3G).

In contrast to the 2250 total proteins identified and 247 proteins differentially represented in female Hjv^−/− vs. Hjv^+/+ bile, only 1231 total proteins were identified and 34 proteins differentially represented in male Hjv^−/− vs. Hjv^+/+ bile (Fig. 4A). Of these, 1 was upregulated and 33 were downregulated in mutant bile (Fig. 4A, B). Fth1 and Ftl1 levels were increased in mutant bile but this did not reach significance, which may reflect the variability in levels of these proteins in male mutant bile as we detected by immunoblot. The iron-binding protein transferrin (Tf) and the hemoglobin-binding protein haptoglobin (Hp) were decreased in mutant bile, but the log fold changes were not prominent at −1.02 for Hp and −1.06 for Tf. Ontology analysis of the 33 proteins less abundant in Hjv^−/− bile did not identify any over-represented pathways. Only eight proteins were differentially abundant in Hjv^−/− vs. Hjv^+/+ bile for both female and male mice (Fig. 4C; proteins labeled * in Fig. 4B).

Fig. 4.

Male Hjv^−/− and Hjv^+/+ mouse bile do not differ prominently in protein content. LC-MS analysis was performed on 30 μg of bile from male Hjv^+/+ and Hjv^−/− mice (n = 5/group). (A) Volcano plot of male Hjv^−/− vs. Hjv^+/+ mouse bile, plotted as in Fig. 3A. (B) Heatmap of proteins differentially represented in Hjv^−/− vs. Hjv^+/+ bile. * indicates proteins differentially abundant in male Hjv^−/− vs. Hjv^+/+ bile and female Hjv^−/− vs. Hjv^+/+ bile. (C) Venn diagram comparing proteins differentially represented in male Hjv^−/− vs. Hjv^+/+ bile and female Hjv^−/− vs. Hjv^+/+ bile.

To explore these sex-specific differences in bile protein content further, we compared the proteomes of female vs. male mice for each genotype. Comparison of female vs. male bile for Hjv^+/+ mice identified 1469 total proteins (Fig. 5A). Of these, 199 proteins were differentially abundant between female and male samples, with 182 more and 17 less abundant in female samples. In addition, 89 (45%) aligned with metabolic pathways (Fig. 5B–D). Comparison of female vs. male bile for Hjv^−/− mice identified 1150 total proteins (Fig. 5E). Of these 1150 proteins, 71 proteins were differentially abundant between female and male samples, with 55 more and 16 less abundant in female samples. Additionally, 24 of these 71 proteins, (34%) aligned with metabolic pathways (Fig. 5Gand H).

Fig. 5.

Female and male mouse bile differ in protein content. (A) Volcano plots of female vs. male Hjv^+/+ mouse bile, plotted as in Fig. 3A. (B and C) Ontology (B) and network (C) analysis of differentially abundant proteins in female vs. male Hjv^+/+ mouse bile. (D) Heatmap of differentially abundant proteins aligning with metabolic pathways in female vs. male Hjv^+/+ mouse bile. (E) Volcano plots of female vs. male Hjv^−/− mouse bile, plotted as in Fig. 3A. (F and G) Ontology (F) and network (G) analysis of differentially abundant proteins in female vs. male Hjv^−/− mouse bile. (H) Heatmap of differentially abundant proteins aligning with metabolic pathways in female vs. male Hjv^−/− mouse bile.

While we are not aware of previous publications documenting sex-specific differences in bile proteomes, there is a precedent for sex-specific differences in bile-related processes. For example, bile acid synthesis and metabolism differ by sex in mice.^43,44 Female sex increases the risk of cholelithiasis or gallstone formation in humans.^45,46 Sex hormones are known to modulate function of cholangiocytes, the cells that line the biliary tree.⁴⁷ Plasma bile acid levels are impacted by sex in mice.⁴⁸ Mice with deficiency in steroid 5-beta-reductase Akr1d1, required for hepatic bile acid synthesis, display sex-specific phenotypes.⁴⁹ Bile acid profiles in intestines, plasma, and liver show sex-specific differences in a mouse model of Alzheimer's disease.⁵⁰ Mice deficient in Fancd2, a component of the Fanconi anemia DNA damage response pathway, develop sex-specific impacts on hepatic bile acid metabolism.⁵¹

As mentioned above, older biochemical and ultrastructural studies implicated lysosomal exocytosis from hepatocytes. The transcription factor TFEB is a key regulator of lysosomal biogenesis, autophagy, and exocytosis. To determine if TFEB plays a role in biliary iron and ferritin excretion, we injected 2-month-old Hjv^−/− mice retro-orbitally with saline or adeno-associated virus 8 (AAV8) carrying shRNA targeting Tfeb, then analyzed them at 4 months of age. AAV treatment decreased Tfeb protein levels but had no impact on holo-ferritin, Ftl1, or Fth1 levels in liver, bile, or serum or non-heme iron levels in liver or bile (Fig. 6). These results suggested that Tfeb is not essential for biliary iron or ferritin excretion.

Fig. 6.

AAV-mediated Tfeb knockdown in Hjv^−/− mice does not impact bile iron or ferritin levels. Hjv^−/− mice were injected retro-orbitally with AAV8 carrying Tfeb shRNA at 2 months of age, then analyzed at 4 months of age. Untreated 4-month-old Hjv^+/+ and Hjv^−/− mice were collected for reference. (A) From top to bottom: Tfeb immunoblots of liver lysates (25 μg); iron staining of native gels of liver lysates (100 μg), bile (10 μl), and serum (1 μl) without (−DAB) and with DAB (+DAB) enhancement. (B) From top to bottom: Ftl1 and Fth1 immunoblots of liver lysate (25 μg), bile (10 μl), and serum (1 μl). (C) Liver non-heme iron levels. (D) Bile non-heme iron levels. For (C and D), at least three samples per group; bars indicate mean ± standard deviation; genotypes compared by one-way analysis of variance (ANOVA) with Tukey's post-hoc test (ns P≥0.05, ^∗P<0.05, ^∗∗P<0.01, ^∗∗∗P<0.001, ^∗∗∗∗P<0.0001).

Beyond the sex-specific differences in bile proteomes that we reported above, our current study, in combination with our published work on bile iron and ferritin levels in mice raised on iron-rich diets, document that systemic iron excess leads to biliary iron excretion in mice. One important facet of biliary iron excretion that remains to be determined is the underlying mechanism. Candidate molecular determinants of this process may be identified from studies of cellular (non-biliary) ferritin secretion, which have highlighted two pathways: secretion of extracellular vesicles and secretory autophagy. These pathways may not be entirely distinct, as they can overlap for other secreted factors.^52,53 Extracellular vesicles are a heterogeneous group of membrane-bound structures that function in waste elimination and communication and exchange of components between cells.⁵⁴ Secretory autophagy is a process in which proteins that lack signal peptides (such as ferritin) are secreted from the cell via an autophagy-based pathway that directs proteins for secretion, not degradation.⁵⁵

Multiple studies of cell lines, model organisms, and patient samples have reported ferritin in extracellular vesicles and exosomes, which are extracellular vesicles of specific size range derived from the endosomal system.^56–71 Two factors have been implicated in cellular ferritin secretion via extracellular vesicles: nuclear receptor activator 4 (NCOA4) and prominin 2 (PROM2). NCOA4 is a ferritin cargo receptor essential for ferritinophagy, a form of autophagy active under conditions of iron demand in which ferritin is transported by NCOA4 to lysosomes for degradation and release of iron stores.⁷² The liver is a key site of NCOA4 expression and function—for example, mice with hepatic Ncoa4 deficiency have impaired mobilization of hepatic iron stores after blood loss.⁷³ A recent study demonstrated that NCOA4 also mediates secretion of ferritin-loaded, CD63-positive extracellular vesicles in iron-loaded fibroblasts, suggesting that NCOA4 also acts in conditions of iron excess.⁶⁰ This study also noted that NCOA4 colocalized with the exosomal marker CD63 and that CD63 expression is regulated by iron regulatory proteins, proteins that post-transcriptionally regulate gene expression in response to changes in iron levels.⁶⁰ The second factor implicated in ferritin secretion via extracellular vesicles is PROM2. This is a membrane protein and putative regulator of lipid dynamics that enhances cancer cell resistance to ferroptosis, an iron-dependent form of cell death, by driving secretion of ferritin-loaded exosomes to decrease iron levels.⁷⁰ It is not yet clear if PROM2-dependent ferritin secretion occurs under physiologic conditions and/or is distinct from NCOA4-dependent ferritin secretion.

In addition to secretion via extracellular vesicles, secretory autophagy has been implicated in cellular ferritin secretion. A study in iron-loaded bone marrow-derived macrophages reported that ferritin can be secreted via secretory autophagy.⁵⁶ This study also analyzed serum ferritin levels in mice with mutations in vesicle trafficking/endolysosomal proteins required for secretory autophagy. Increased serum ferritin and liver iron levels and decreased serum iron levels were noted in mice carrying the ashen splice site loss-of-function mutation in Rab27a, a Rab GTPase essential for vesicle transport including exosome secretion. Altered serum ferritin and liver iron levels were also noted in mice carrying the palladin/BLOC-1 mutation in Bloc1s6, the ruby eye/BLOC-2 mutation in Hps6, and the pale ear/BLOC-3 mutation in Hps1—all of these genes encode BLOC (biogenesis of lysosomal organelle complex) subunits. It is not clear if these mouse phenotypes reflect a direct role for Rab27a, Bloc1s6, Hps6, and Hps1 in cellular ferritin secretion. However, a separate study of non-alcoholic fatty liver disease/steatohepatitis showed that Rab27a mediates the secretion of iron-loaded extracellular vesicles from hepatocytes with subsequent uptake of vesicles by hepatic stellate cells, leading to iron excess, oxidative stress, and fibrogenesis.⁷⁴ Additionally, a recent study reported that knockdown of Rab27a in iron-loaded primary astrocytes enhanced ferritin secretion.⁷⁵

Our future investigations will focus on determining the impact of candidate determinants on biliary iron excretion. Once the mechanism of biliary excretion of excess iron is established, the relevance of this process to iron homeostasis can be directly interrogated. The fact that excessive dietary iron absorption in hereditary hemochromatosis leads to iron excess suggests that biliary iron excretion cannot compensate for increased iron uptake in this disease. However, pharmacologic stimulation of biliary iron excretion could serve as a useful adjunct to pharmacologic inhibition of dietary iron absorption in diseases of iron excess.

Conclusions

Like mice raised on iron-rich diets, hemojuvelin-deficient mice have increased bile iron and ferritin levels. Bile proteomes differed between female and male mice for both wild-type and hemojuvelin-deficient mice, suggesting sex and iron excess both impact bile protein content. Treating mice with AAVs carrying Tfeb shRNA decreased liver Tfeb levels but had no impact on bile iron or ferritin levels. These results can serve as the basis for future studies investigating the molecular basis and physiologic relevance of biliary iron excretion.

Author contributions

Study design: M.P., T.B.B.; execution of experiments: M.P., L.C., J.Z.Z., G.S.C., N.D., T.B.; data analysis: M.P., N.A.D., T.B.B.; manuscript writing: T.B.B.; manuscript revision: all authors.

Funding

This work was supported by National Institutes of Health (NIH) R01 DK110049 to TBB. Proteomic studies were conducted using the Mass Spectrometry Proteomics Core Facility (formerly Rhode Island NSF/EPSCoR Proteomics Shared Resource Facility), which was supported in part by the National Science Foundation EPSCoR Grant No. 1004057, National Institutes of Health Grant No. 1S10RR020923, S10RR027027 (Orbitrap XL ETD Mass Spectrometer), a Rhode Island Science and Technology Advisory Council grant, and the Division of Biology and Medicine, Brown University.

Conflicts of interest

The authors have no conflicts of interest to declare.

Data availability

Proteomics data for this article are included as online supplementary material. Uncropped blot and gel images and all data are available upon request.