Implication of the proprotein convertases in iron homeostasis: Proprotein convertase 7 sheds human transferrin receptor 1 and furin activates hepcidin


  • Johann Guillemot,

    1. Laboratory of Biochemical Neuroendocrinology, Clinical Research Institute of Montreal (IRCM), Affiliated to the University of Montreal, Montreal, Quebec, Canada
    Search for more papers by this author
  • Maryssa Canuel,

    1. Laboratory of Biochemical Neuroendocrinology, Clinical Research Institute of Montreal (IRCM), Affiliated to the University of Montreal, Montreal, Quebec, Canada
    Search for more papers by this author
  • Rachid Essalmani,

    1. Laboratory of Biochemical Neuroendocrinology, Clinical Research Institute of Montreal (IRCM), Affiliated to the University of Montreal, Montreal, Quebec, Canada
    Search for more papers by this author
  • Annik Prat,

    1. Laboratory of Biochemical Neuroendocrinology, Clinical Research Institute of Montreal (IRCM), Affiliated to the University of Montreal, Montreal, Quebec, Canada
    Search for more papers by this author
  • Nabil G. Seidah

    Corresponding author
    1. Laboratory of Biochemical Neuroendocrinology, Clinical Research Institute of Montreal (IRCM), Affiliated to the University of Montreal, Montreal, Quebec, Canada
    • Laboratory of Biochemical Neuroendocrinology, Clinical Research Institute of Montreal, 110 Pine Ave West, Montreal, Quebec, H2W 1R7, Canada
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

  • Supported by CIHR grant MOP-44363, a Strauss Foundation grant, and a Canada Chair #216684. J.G. is a postdoctoral fellow supported by the Fonds de la Recherche en Santé du Québec (FRSQ).


The first seven members of the proprotein convertase (PC) family activate protein precursors by cleavage after basic residues. While PC7 has no known specific substrates, it shows redundancy with other PCs. A genome-wide association study suggested that circulating levels of shed human transferrin receptor 1 (hTfR1) are regulated by PC7. We thus examined whether hTfR1 constitutes a specific substrate for PC7. Coexpression of hTfR1 with PCs in several cell lines indicated that PC7 is the only convertase that sheds this receptor into the medium. Site-directed mutagenesis showed that cleavage occurs at the unusual site KTECER100↓LA, in which the P1 Arg100 and P6 Lys95 are critical. Pharmacological treatments revealed that shedding of hTfR1 by PC7 requires endocytosis into acidic clathrin-coated vesicles. A PC7 chimera, in which the transmembrane domain and the cytosolic tail of PC7 were replaced by that of the convertase furin, lost its ability to cleave the receptor, demonstrating the importance of these domains in the regulation of PC7 function. Analysis of primary hepatocytes from mice lacking furin, PC5, PACE4, or PC7 revealed that hepcidin, which limits iron availability in the circulation, is specifically generated by furin and not by PC7. Finally, depletion of iron in the medium of hepatoma cell lines incubated with the iron chelator desferrioxamine resulted in PC7 down-regulation. Conclusion: Among the PC family members, only furin activates hepcidin in hepatocytes, and uniquely the full-length membrane-bound PC7 can directly shed hTfR1 by cleavage at Arg100↓. Our results support the notion that, when iron is limiting, hTfR1 levels increase at least in part by way of the down-regulation of PC7 expression. (HEPATOLOGY 2013;)

Many secretory proteins are synthesized as precursors that are activated by limited proteolysis at specific sites by the proprotein convertases (PCs), a family of nine secretory serine proteases related to bacterial subtilisin and yeast kexin (genes PCSK1 to PCSK9).1 PC7, a type-I membrane-bound protease that is ubiquitously expressed,2 is the most ancient and highly conserved member of the PC family. It belongs to a subgroup of seven PCs that cleave their substrates after basic residues, often organized as doublets.1 Like all PCs, it is synthesized as a precursor, which upon folding in the endoplasmic reticulum undergoes autocatalytic cleavage of its N-terminal inhibitory prosegment.2 Other posttranslational modifications include N-glycosylation, sulfation, and palmitoylation of the cytosolic Cys699 and Cys704.2, 3 Like furin, PC7 accumulates in the trans Golgi network, but also cycles to the cell surface and back to the trans Golgi network by way of endosomes.4, 5 Internalization of surface PC7 into clathrin-coated vesicles is dependent on the presence of a 14 amino acids (aa) stretch in its 97 aa-long cytosolic tail (CT), which contains a key Pro-Leu-Cys726 motif.5 Importantly, and different from furin, PC7 is never found as a soluble shed form in the medium,6 indicating that its activity is intracellular or linked to the cell surface. The physiological roles of PC7 are largely unknown and no PC7-specific substrate has yet been identified. We reported that PC7 is the only PC that indirectly enhances the processing of the cell surface pro-epidermal growth factor (proEGF) through the activation of a latent protease.7

A recent genome-wide association study8 established a strong link between plasma levels of the soluble human transferrin receptor 1 (s-hTfR1) and a single nucleotide polymorphism (SNP) in intron 9 of the human PC7 gene (PCSK7). The authors suggested that PC7 could influence iron homeostasis, either by direct shedding of the hTfR1 or indirectly by way of the activation of hepcidin, the major regulator of iron homeostasis.9 Human TfR1 is a homodimeric type II membrane protein that mediates cellular uptake of iron.10 Although the role of s-hTfR1 remains unclear, the measurement of its plasma levels is a valuable tool to distinguish anemia due to inflammation (unchanged levels) from anemia due to inflammation combined with iron deficiency (resulting in higher levels of s-hTfR1).11 Dietary iron is absorbed by enterocytes and exported by way of the iron channel ferroportin into plasma, whereupon it binds to transferrin. The complex is internalized in cells, primarily reticulocytes, by hTfR1 binding and subsequent erythrocyte phagocytosis by macrophages results in iron recycling into the plasma. Iron homeostasis is tightly regulated by the liver. Under conditions of excess iron, hepcidin is secreted primarily by hepatocytes and limits the amount of circulating iron, as it blocks iron release by enterocytes and macrophages by enhancing the degradation of the iron channel ferroportin,9 while hTfR1 expression is down-regulated. Conversely, upon iron depletion, hepcidin is down-regulated9 and hTfR1 is up-regulated.12 In the present study we demonstrated that PC7 is the only convertase that sheds hTfR1, whereas furin is the convertase that generates mature hepcidin in hepatocytes. Moreover, incubation of hepatoma cells under iron-deficient conditions known to result in the up-regulation of hTfR1 levels also leads to a reduction of PC7 levels, likely as a compensatory mechanism to enhance cellular iron uptake.

Materials and Methods

Cell Culture, Transfections, and Cell Treatments.

HuH7, HepG2, HEK293, and COS-1 cell lines were grown in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) with 10% fetal bovine serum (FBS, Invitrogen), CHO-K1, and CHO ldlD cell lines were grown in DMEM/F12 medium with 10% FBS, whereas the K562 cells were grown in Roswell Park Memorial Institute (RPMI) 1640 medium (Invitrogen) with 10% FBS. All cells were maintained at 37°C under 5% CO2. HuH7, COS-1, CHO-K1, CHO ldlD, and K562 cells were cotransfected with equimolar quantities (∼0.8 μg) of each plasmid using Lipofectamine (Invitrogen), HepG2 cells with FuGENE HD (Roche Diagnostics) and HEK293 cells with Effectene (Qiagen) according to the manufacturer's instructions. At 24 hours posttransfection, cells were washed in serum-free medium followed by incubation for an additional 20 hours in serum-free medium alone or in combination with 25 μM GM6001 (Calbiochem), 20 μM TAPI-2) (Calbiochem), 25 μM (2S,3S)-trans-epoxysuccinyl-L-leucylamido-3-methylbutane ethyl ester (E64d; Calbiochem), 300 μM serine protease inhibitor AEBSF (Roche Applied Science), 0.5 mg/mL α1-Antitrypsin (α1-AT; Calbiochem), 50 μM decanoyl-RVKR-chloromethyl ketone (RVKR-cmk; Bachem), 10 μM of the hexapeptide (D-Arg)6 (D6R; EMD Chemicals), 2.5 μg/mL brefeldin A (BFA; Calbiochem), 5 mM ammonium chloride (NH4Cl; Sigma), 1 μg/mL filipin III (Sigma), or 1 mM methyl-β-cyclodextrin (MβCD; Sigma). Sucrose (0.4M; Sigma) and dynasore (80 μM; Sigma) were added during 1-hour posttransfection. Following treatments, cells and conditioned media were collected for western blot analysis. For total RNA and protein isolation, HuH7 cells were incubated with 100, 200, or 400 μg/mL ferric ammonium citrate (FAC; Sigma) or 100, 200, 400 μM desferrioxamine (DFO; Santa Cruz Biotechnology) for 24 hours and collected by way of standard methods.


aa, amino acid; ADAM, a disintegrin and metalloprotease; CT, cytosolic tail; DFO, desferrioxamine; hTfR, human transferrin receptor; mRNA, messenger RNA; mTfR, mouse TfR; PACE4, paired basic amino acid cleaving enzyme 4; proEGF, pro-epidermal growth factor; PC(s), proprotein convertase(s); PCSK, proprotein convertase subtilisin/kexin; s-hTfR, soluble hTfR; SNP, single nucleotide polymorphism; TfR, transferrin receptor; TM, transmembrane domain; WT, wildtype.

Isolation, Culture, and Transfection of Mouse Primary Hepatocytes.

Mouse primary hepatocytes were isolated from 8 to 10-week-old male livers using a two-step collagenase perfusion method, as described.13 In 3.5-cm Petri dishes coated with fibronectin (0.5 mg/mL, Sigma), 5 × 105 cells were seeded in Williams' medium E (Invitrogen) with 10% FBS. After 2 hours the medium was replaced with hepatozyme medium (Invitrogen) for 12 hours prior to transfection. Transfections were performed with Effectene using a total of 4 μg of complementary DNA (cDNA), following the manufacturer's instructions. Cell lysates and media were collected 48 hours posttransfection and subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) separation (14% Tris-Tricine) followed by western blot analysis.

For a description of the rest of the experimental procedures, see the Supporting Material.


PC7 Sheds hTfR1 by Cleavage at Arg100.

To assess the shedding of hTfR1 by PCs, we coexpressed hTfR1 carrying a V5-tag at its C-terminus (Fig. 1A) with the basic-aa specific convertases PC1, PC2 (with its chaperone 7B2), furin, PC5A, PC5B, paired basic aa cleaving enzyme 4 (PACE4), or PC71 in human hepatoma HuH7 and HepG2 cells (Fig. 1B,C), human embryonic kidney HEK293, and monkey kidney COS-1 cells (Supporting Fig. 1A,B). In all cases, only PC7 generated an ∼85 kDa product that has a similar molecular mass to circulating hTfR1 in human serum,14 suggesting that hTfR1 is shed by PC7. Coexpression of hTfR1 with PC7 in the human promyelocytic leukemia K562 cells and Chinese hamster ovarian (CHO)-K1 cells gave the same results (Supporting Fig. 1C,D). In HuH7 cells, the ∼85 kDa fragment was also detected in cell lysates (Fig. 1B,D). Finally, because PC7 itself is never shed into the medium in any cell line,4, 6 it was not surprising that a soluble form of PC7 could not process hTfR1 (Fig. 1D), although soluble PC7 was active in in vitro enzymatic assays.6 In this study we mainly used HuH7 cells, as they express high levels of endogenous PC7 and hTfR1 (Supporting Fig. 2). In HEK293 and COS-1 cells, an ∼75 kDa form was also detected in the media, even in the absence of any PC (Supporting Fig. 1A,B), likely due to an endogenous protease (not blocked by the metalloprotease inhibitor GM6001; data not shown) cleaving hTfR1 at a further downstream site.

Figure 1.

hTfR1 is shed only by PC7. (A) Schematic representation of the 760-aa hTfR1 and its soluble form. Depicted are the cytosolic tail (CT), the transmembrane domain (TM), the predicted PC-processing site (KTECER100↓), and the C-terminal V5-tag. Western blot analysis of lysates and 20-hour conditioned media from HuH7 (B) and HepG2 (C) cells expressing hTfR1-V5 and either empty pIRES-EGFP vector or diverse PCs, or (D) HuH7 cells expressing hTfR1-V5 and either full-length PC7 or its soluble form (sPC7). The separated proteins were revealed using mAb V5-HRP and loading was estimated with a β-actin antibody. This result is representative of at least two separate experiments.

Because PC7 cleaves substrates after basic residues,1 we mutated the basic aa in the proposed cleavage site KTECER100↓ of hTfR1,14 as well as in two downstream putative sites, ArgArg121 and LysArg129 (Fig. 2A). Coexpression of PC7 with wildtype (WT) hTfR1 or its mutants K95A, R100A, R100E, or R100K revealed that the latter were resistant to cleavage, whereas the R121A and R129A mutants were well processed in HuH7 (Fig. 2B) and COS-1 (Supporting Fig. 3A) cells. This strongly suggests that shedding of hTfR1 by PC7 occurs by way of cleavage at Arg100↓, and that the recognition site KTECER100↓LA contains a critical P1 Arg and P6 Lys within the motif KXXXXR↓. In COS-1 cells, we estimate that, after a 20-hour incubation, the amount of s-hTfR1 in the medium represented ∼50% that of cellular hTfR1 (Supporting Fig. 3B). Active cell surface hTfR1 is a disulfide-linked homodimer of monomeric chains in which Cys89 and Cys98 in one chain are attached to the same residues in the other chain (Fig. 2A).15 Unusually, Cys98 occupies the P3 position of the shedding site. This prompted us to assess whether these disulfide bridges had any impact on the ability of PC7 to shed hTfR1. In HuH7 cells, coexpression of PC7 with WT hTfR1 (dimer), or its mutants C89A, C98A (mixture of monomer and dimer), or the monomeric [C89A,C98A], demonstrated that WT and hTfR1 mutants were all shed by PC7 and that s-hTfR1 generation is enhanced for hTfR1 mutated at Cys98 (Fig. 2C). Thus, PC7 cleaves the physiological dimeric and the monomeric forms of hTfR1. However, the absence of one of the disulfide bridges seems to reduce the efficacy of disulfide bond formation of the other, since monomeric forms are now present in the C89A or C98A mutants. The absence of monomeric forms in WT hTFR1 favors the hypothesis that PC7 essentially cleaves the dimeric form.

Figure 2.

Identification of the KTECER100↓LA cleavage site by site-directed mutagenesis. (A) Sequence of hTfR1. Potential cleavage sites are in gray boxes and mutated amino acids are in bold. Homodimers are linked by two disulfide bonds. (B,C) Cell lysates and 20-hour conditioned media from HuH7 cells expressing hTfR1-V5 carrying either no mutation (WT) or (B) K95A, R100A, R100E, R100K, R121A, and R129A mutations or (C) C89A, C98A, [C89A,C98A] mutations, and either empty vector or PC7 were analyzed by western blotting using a mAb V5-HRP or a β-actin antibody. This result is representative of at least two separate experiments.

No s-hTfR1 could be detected in the media in the absence of overexpressed PC7. To enhance endogenous shedding, we exploited the fact that O-glycosylation at Thr104 partially inhibits hTfR1 shedding.16 In agreement, upon overexpression of a T104D mutant of hTfR1, we were able to detect the non-O-glycosylated shed form of the T104D mutant, even in the absence of PC7 (Fig. 3A) and the general PC inhibitor, decanoyl-RVKR-cmk reduces s-hTfR1 shedding (Fig. 3B). This supports the notion that endogenous PC7 in HuH7 cells can shed hTfR1. To strengthen this conclusion, we coexpressed the T104D mutant with the prosegment of PC7, a PC7-specific inhibitor. This resulted in a marked inhibition (60%) of the shedding of the T104D mutant (Fig. 3C). Moreover, as previously demonstrated,17 in the ldlD CHO cell line defective in O-glycosylation, the endogenous shedding of WT hTfR1 was enhanced, while it was inhibited (58%) upon hTfR1 and PC7 prosegment coexpression (Fig. 3D). We conclude that the membrane-bound PC7 is the cognate sheddase of hTfR1.

Figure 3.

Inhibition of hTfR1 shedding by blockage of endogenous PC7 activity. (A) Cell lysates and 20-hour conditioned media from HuH7 cells expressing hTfR1-V5 carrying either no mutation (WT), or a T104D mutation. (B) Cell lysates and conditioned media from COS1 cells, expressing hTfR1 (T104D)-V5 mutant collected after no treatment (−) or treatment (+) with the cell general PC-inhibitor decanoyl-RVKR-cmk (RVKR-cmk, 25 μM). (C) Representative western blot and relative amounts of s-hTfR1/hTfR1 ratio of cell lysates and 20-hour conditioned media from HuH7 cells expressing the hTfR1(T104D)-V5 mutant and either the inhibitory prosegment of PC7 (pPC7), vector alone, or WT PC7. (D) Representative western blot and relative amounts of s-hTfR1/hTfR1 ratio of cell lysates and 20-hour conditioned media from CHO ldlD cells expressing hTfR1-V5 and either vector alone or pPC7. The separated proteins were revealed using mAb V5-HRP or a β-actin antibody. This result is representative of at least three separate experiments.

PC7 Does Not Shed Mouse TfR1.

With the objective of testing the implication of PC7 in iron metabolism in mice lacking this convertase, we first verified whether PC7 sheds mouse TfR1 (mTfR1). However, coexpression of mTfR1-V5 in HuH7 cells with PC1, PC2/7B2, furin, PC5A, PC5B, PACE4, or PC7,1 revealed that no PC can shed the mouse protein (Fig. 4A). Similar results were obtained in COS-1 cells (data not shown). Alignment of the mTfR1, rTfR1 and hTfR1 sequences revealed a Lys at the putative P1 cleavage site of the mouse protein instead of a P1 Arg in human or rat TfR1 (Fig. 4B). To test the importance of the P1 Arg for processing by PC7, we coexpressed PC7 with a K100R mutant of mTfR1. This resulted in the secretion of a soluble ∼85 kDa mTfR1 into the medium (Fig. 4C). Conversely, the hTfR1 R100K mutant was not cleaved by PC7 (Fig. 2A). We conclude that the inability of PC7 to shed mTfR1 is due to the absence of the critical P1 Arg and that likely rat, but not mouse, TfR1 can be shed by PC7.

Figure 4.

Different from human, PC7 does not shed mouse TfR1 (mTfR1), but could shed rat TfR1. (A) Western blot analysis of lysates and 20-hour conditioned media from cells expressing mTfR1-V5 and either empty vector or diverse PCs in HuH7 cells. (B) Comparison of human (h), mouse (m), and rat (r) TfR1 sequences. (C) Cell lysates and 20-hour conditioned media from HuH7 cells expressing mTfR1-V5 carrying either no mutation (WT) or a K100R mutation (mimicking the rat sequence) and either empty vector or PC7, were analyzed by western blotting. The separated proteins were revealed using mAb V5-HRP or a β-actin antibody. This result is representative of at least two separate experiments.

Membrane-Bound PC7 Directly Enhances the Shedding of hTfR1.

Although the cleavage of hTfR1 after a P6 Lys and P1 Arg (KXXXXR↓) suggests that is directly achieved by a PC,1 we verified that this cleavage was not performed indirectly by way of the activation of a latent protease, as for proEGF processing.7 Cleavage of proEGF by PC7 was blocked by the serine protease inhibitors AEBSF and α1-AT, as well as by the cell-permeable Cys-protease inhibitor E64d.7 However, shedding of hTfR1 was insensitive to these reagents (Supporting Fig. 4A). Metalloproteases, such as ADAM (for a disintegrin and metalloprotease) family members, are implicated in the shedding of many cell surface proteins18 and, based on preparations of membranes from cell lines, were reported to participate in the constitutive shedding of hTfR1.19 However, again, neither the general metalloprotease inhibitor GM6001 nor the ADAM inhibitor TAPI-2 had any effect on the shedding of hTfR1 by PC7 in HuH7 cells (Supporting Fig. 4B) or COS-1 cells (data not shown). This suggests that, even though PC7 can activate ADAM-10, which is in turn implicated in the processing of β-amyloid precursor,20 it does not activate a metalloprotease responsible for hTfR1 shedding in the cell lines tested.

Finally, because the genome-wide association study also indicated an association between circulating hTfR1 levels and a coding SNP in the gene of matriptase-2,8 which is implicated in iron metabolism,21 we evaluated whether this serine protease constitutes an intermediate enzyme activated by PC7. Coexpression of hTfR1 with matriptase-2 in the absence or presence of PC7 in HuH7 cells revealed that matriptase-2 overexpression did not result in hTfR1 shedding, nor modified the PC7 activity on hTfR1 (Supporting Fig. 4C). Altogether, these data suggest that membrane-bound PC7 is directly responsible for the shedding of hTfR1.

Shedding Occurs in Endosomes.

We and others had already reported that PC7 is localized in the trans Golgi network, endosomes, and at the cell surface (Supporting Fig. 5A,B).4, 5 Similarly, hTfR1 is also localized at the cell surface and endosomal-like vesicular structures of HuH7 cells (Supporting Fig. 5C,D). However, upon coexpression of the two proteins we noticed that, under nonpermeabilizing conditions, hTfR1 and PC7 can colocalize at the cell surface (Supporting Fig. 5E) and, under permeabilizing conditions, within intracellular trans Golgi network- and vesicular-like compartments (Supporting Fig. 5F). These data support the hypothesis that PC7 colocalizes with hTfR1, and hence is well poised to directly shed this receptor.

The use of two general PC inhibitors, the cell-permeable decanoyl-RVKR-cmk and noncell-permeable hexapeptide (D-Arg)6 revealed that shedding does not occur at the cell surface (Fig. 5A). In HuH7 cells, the cell-permeable inhibitor resulted in a drastic reduction of s-hTfR1 in the media and cells, whereas the cell surface-specific inhibitor had no effect. Furthermore, incubation of HuH7 cells with brefeldin A, which inhibits endoplasmic reticulum to Golgi transport, prevented the secretion and generation of s-hTfR1 (Fig. 5B), demonstrating that processing of hTfR1 by PC7 occurs in a postendoplasmic reticulum compartment.

Figure 5.

hTfR1 shedding occurs in endosomes. Western blot analysis of lysates and conditioned media from HuH7 cells, expressing hTfR1-V5 and either empty vector or PC7, collected after no treatment (−) or treatment (+) with either (A) the cell permeable PC-inhibitor decanoyl-RVKR-cmk (RVKR-cmk, 25 μM) or the cell surface PC-inhibitor hexapeptide (D-Arg)6 (D6R, 10 μM), (B) brefeldin A (BFA, 2.5 μg/mL), (C) filipin III (1 μg/mL) or methyl-β-cyclodextrin (MβCD, 1 mM), (D) ammonium chloride (NH4Cl, 5 mM), and (E) sucrose (0.4 M) or dynasore (80 μM). The separated proteins were revealed using mAb V5-HRP or a β-actin antibody. This result is representative of at least three separate experiments.

Because the enhancement of proEGF cleavage by PC7 was shown to be neutralized by methyl-β-cyclodextrin, which disrupts cholesterol-rich microdomains, we treated HuH7 cells that coexpressed hTfR1 and PC7 with this drug (1 mM) or with filipin III (1 μg/mL) that inhibits lipid raft- and caveolae-mediated endocytosis. However, both drugs failed to block hTfR1 shedding (Fig. 5C). To further identify the compartment in which cleavage takes place, HuH7 cells were treated with 5 mM NH4Cl, which increases the pH of acidic compartments. This inhibited the shedding of hTfR1 (Fig. 5D), suggesting that processing occurs in an acidic compartment. Furthermore, incubation of cells with hypertonic sucrose (0.4 M) that induces dispersion of clathrin lattices on the plasma membrane, or with dynasore (80 μM), a cell-permeable inhibitor of dynamin GTPase activity that blocks clathrin-dependent endocytosis, also prevented the generation of s-hTfR1 by PC7 (Fig. 5E). These data agree with a report showing that hTfR1 internalization occurs by way of a clathrin-dependent mechanism inhibited by sucrose, whereas filipin and methyl-β-cyclodextrin had no effect.22 Thus, hTfR1 shedding by PC7 requires endocytosis of hTfR1 into acidic clathrin-coated vesicles.

Transmembrane Domain and Cytosolic Tail of PC7 Are Critical for hTfR1 Shedding.

Because soluble PC7 cannot shed hTfR1, we examined the requirement for the transmembrane domain (TM) alone, or both the TM-cytosolic tail (TMCT) domains of PC7 in hTfR1 shedding. For this, we swapped the above domains of human PC7 with the corresponding ones in human furin (PC7CT-furin and PC7TMCT-furin) and vice versa (furinCT-PC7 and furinTMCT-PC7) (Fig. 6A). Replacement of the PC7 CT by that of furin appreciably reduced hTfR1 shedding, whereas replacement of the PC7 TMCT by that of furin completely abolished the shedding activity of PC7 (Fig. 6B), indicating that both domains are important for hTfR1 shedding. In agreement, none of the furin chimeras could process hTfR1. Since the above chimeras used could all process proEGF,7 this clearly distinguishes the direct PC7 processing of hTfR1 from the indirect proEGF cleavage.

Figure 6.

The transmembrane-cytosolic tail (TMCT) of PC7 is critical for hTfR1 shedding. (A) Schematic diagrams depicting chimeras of PC7 and furin, also indicating the palmitoylated Cys of PC7 in its CT. (B) Western blot analysis of lysates and 20-hour conditioned media from HuH7 cells expressing hTfR1-V5 and either empty vector, PC7, its chimeras PC7CT-furin, PC7TMCT-furin, furin, or its chimeras furinCT-PC7, furinTMCT-PC7. (C,D) Western blot analysis of lysates and 20-hour conditioned media from HuH7 cells expressing hTfR1-V5 and either PC7 (WT), or the non-palmitoylated PC7-[C699A,C704A] mutant (C) or PC7-(D188G), PC7-(R316C) and PC7-(fsL506P) mutants (D). The separated proteins were revealed using mAb V5-HRP, PC7 or β-actin antibodies. This result is representative of at least three separate experiments.

PC7 is the only convertase that has palmitoylated cysteines (Cys699 and Cys704) within its CT (Fig. 6A).4, 23 However, these posttranslational modifications are not required for hTfR1 shedding, as the C699A and C704A mutants were as efficient as WT PC7 to cleave the receptor (Fig. 6C). Finally, upon examination of the 41 missense SNPs reported in the human PCSK7 gene database ( we anticipated that three of them should be deleterious: D188G, R316C, and frameshift fsL506P. Coexpression of the corresponding mutants with hTfR1 in HuH7 cells led to proPC7 accumulation, indicating that autocatalytic processing of the PC7 zymogen was severely impaired, and to complete loss of hTfR1 shedding (Fig. 6D).

Furin Activates Hepcidin in Primary Hepatocytes.

In an effort to further define the role of PC7 in iron homeostasis, we assessed its ability to activate the central iron regulator hepcidin.9 Previous experiments in cell lines revealed that processing of the inactive precursor human pro-hepcidin into active hepcidin was redundantly achieved by the overexpressed furin, PACE4, PC5, and PC7.24 The availability of knockout mice, in which these convertases are no longer expressed in hepatocytes, allowed us to examine their respective implication in pro-hepcidin processing in a more physiological setting. Pro-hepcidin carrying a V5-tag at its C-terminus (Fig. 7A) was expressed in Furin−/−, PC5−/−, [Furin−/−, PC5−/−] (double knockout), PACE4−/−, and PC7−/− primary hepatocytes. Western blot analyses of cell lysates and media revealed that furin was the only endogenous convertase that was essential for pro-hepcidin processing and secretion of mature ∼4.5 kDa hepcidin (Fig. 7B). Furin seems thus to be the only physiological convertase that generates mature hepcidin in hepatocytes.

Figure 7.

In hepatocytes pro-hepcidin is only shed by furin. (A) Schematic representation of the 84-aa h-pre-pro-hepcidin and its hepcidin mature product. Depicted are the signal peptide (SP), the PC-processing site (RRRRR59↓) and the C-terminal V5 tag. (B) Western blot analysis of lysates and 48-hour conditioned media from primary hepatocytes isolated from WT, PC7−/−, PACE4−/−, PC5−/−, Furin−/− mice and from those lacking both Furin and PC5 [Furin−/−, PC5−/−] expressing h-pro-hepcidin-V5. The separated proteins were revealed using mAb V5-HRP. This result is representative of at least two separate experiments.

PC7 Expression Is Down-regulated by Iron Depletion.

Iron is known to modulate the messenger RNA (mRNA) levels of hTfR1.10 Accordingly, following a 24-hour incubation with an excess of ferric ammonium citrate, HuH7 cells exhibited up to 3.5-fold lower levels of hTfR1 in a dose-dependent fashion. Conversely, when these cells were incubated with the iron chelator desferrioxamine (DFO), the levels of hTfR1 transcripts were up to 2.8-fold higher (Fig. 8A). Under the same conditions, PC7 mRNA levels were not affected by an iron excess, but were ∼2-fold reduced by DFO (Fig. 8B). In agreement, PC7 protein levels were unaffected by 200 μg/mL of ferric ammonium citrate, but significantly reduced (>2-fold) by 200 μM of DFO (Fig. 8C,D). In the latter conditions, the analysis of the mRNA levels of others PCs revealed that excess iron results in a slight down-regulation of furin (Fig. 8E), but that PC7 was the only PC to be down-regulated by 200 μM DFO (Fig. 8F). Finally, we obtained similar results for PC7, hTfR1, and other PCs expression in HepG2 cells (Supporting Fig. 6) and K562 cells (Supporting Fig. 7).

Figure 8.

Effect of ferric ammonium citrate (FAC) or desferrioxamine (DFO) treatments on PCs and hTfR1 levels in HuH7 cells. hTfR1 (A) and PC7 (B) mRNA levels were measured by quantitative polymerase chain reaction from HuH7 cells treated without (0) or with FAC (100, 200, or 400 μg/mL) or DFO (100, 200, or 400 μM). Representative western blot and relative amounts of hTfR1 (C) and PC7 (D) of HuH7 cells treated without (−) or with FAC (200 μg/mL) or DFO (200 μM). The separated proteins were revealed using PC7, hTfR1, or β-actin antibodies. Comparative effect of 200 μg/mL FAC (E) or 200 μM DFO (F) treatments on hTfR1, hFurin, hPC5, hPACE4, hPC7 mRNA levels. Results are expressed as mean values ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. This result is representative of at least three separate experiments.


Our work demonstrates that PC7 can directly shed hTfR1 in cell lines, supporting a genome-wide association study that positively correlated the SNP rs236918 in PCSK7 to higher circulating hTfR1 levels.8 Being intronic and relatively common, the rs236918 polymorphism shows a G-to-C change (minor C allele frequency = 23.8%). Another SNP, rs855791, which is responsible for the A736V substitution in human matriptase-2, was also associated, but weakly, with the levels of s-hTfR1.8 However, our data showed that matriptase-2 did not shed hTfR1 in HuH7 cells (Supporting Fig. 4C). Previous studies also suggested that ADAM metalloproteases are the physiological sheddase(s) of hTfR1.19, 25 The ability of the general PC-inhibitor decanoyl-RVKR-cmk to block shedding was used as an argument to suggest that a furin-like convertase activated the cognate ADAM-related sheddase(s). Our data clearly showed that PC7 is a direct sheddase of hTfR1, without the participation of ADAMs (Supporting Fig. 4B). In addition, we observed a minor and smaller ∼75 kDa shed product in HEK293 and COS-1 cells (Supporting Fig. 1A,B). However, the generation of this ∼75 kDa fragment was not inhibited by general metalloprotease (GM6001; data not shown) or PC (decanoyl-RVKR-cmk) inhibitors. Thus, we cannot exclude the participation of another sheddase in the generation of soluble hTfR1, as the existence of a shedding site distinct from that of PC7 at Arg100↓ was also reported.25, 26

We demonstrated that shedding of hTfR1 does not occur at the cell surface, but rather in acidic endosomes (Fig. 5), in agreement with Rutledge et al.,27 who suggested that hTfR1 shedding was mainly localized in an endosomal compartment. Human TfR1 constitutes the first physiological substrate of PC7. The unique ability of PC7 among PCs to cleave this substrate may reside in the unusual processing site KTECER100↓ that lacks the presence of either a doublet of basic residues at P2-P1 or a P4 Arg, elements known to greatly favor PC cleavage.1 While the presence of a P3 Cys that forms a disulfide bridge upon hTfR1 dimerization15 is unusual in a consensus PC site,28 dimeric hTfR1 is shed by PC7 even though hTfR1 with a free P3 Cys is more efficiently cleavable (Fig. 2C).

PCs were first implicated in iron homeostasis based on their ability to process pro-hepcidin. First, furin knockdown in HepG2 cells or human primary hepatocytes led to reduced levels of hepcidin29 and, second, coexpression of pro-hepcidin with more than one basic aa-specific PC generated active hepcidin.24, 30 On the other hand, PCs may inhibit hepcidin expression through shedding of hemojuvelin that acts as an antagonist in the BMP/Smad pathway responsible for hepcidin mRNA up-regulation.31 Finally, BMPs were shown to be activated by PCs (essentially by furin, PC5, and PACE4).1 Taking advantage of our mouse models, we now show that in primary hepatocytes furin, but not the other PCs, is responsible for pro-hepcidin processing (Supporting Fig. 8).

This study unambiguously demonstrates the ability of PC7 to shed hTfR1, preferring its non-O-glycosylated form. Moreover, under iron depletion in hepatoma cells, PC7 mRNA and protein levels were reduced, whereas those of hTfR1 were increased, suggesting that reduced cleavage of hTfR1 by PC7 further enhances cellular iron uptake (Supporting Fig. 8). However, no 5′ iron responsive element was found in PC7 mRNA. Indeed, low iron levels induce the binding of iron regulatory proteins to these 5′ elements, thereby blocking mRNA translation.12 Low iron levels are also known to regulate target genes by way of binding of hypoxia-induced factor 1α to hypoxia-responsive elements.32 However, such element(s) were also absent from the PCSK7 gene.33 Thus, PC7 down-regulation by iron-depletion in hepatocyte-derived and promyelocytic K562 cells occurs by an as yet undefined mechanism. The results obtained in K562 cells suggest that PC7 is probably regulated by iron status in erythrocyte precursors, which constitute the richest source of hTfR1 and its soluble circulating form. Studies that will examine the correlation between PC7 expression and circulating hTfR1 levels should be carefully scrutinized.

Could the modulation of hTfR1 shedding be associated with disease(s)? (1) hTfR1 can be the receptor of arenaviruses that cause hemorrhagic fevers. During the acute-phase response to infection, TfR1 is up-regulated as a result of iron sequestration, leading to increased virus entry, while s-hTfR1 inhibits the infection of 293T cells by arenaviruses.34 Therefore, it is possible that sequestration of circulating iron may increase hTfR1 and reduce PC7 levels, favoring viral infection. (2) Ethanol increases hTfR1 expression and concomitantly decreases s-hTfR1 in human hepatoma HepaRG cells.35 Because iron metabolism is dysregulated in alcoholic liver diseases, it will be worthwhile to examine the contribution of PC7 in the reduction of s-hTfR1, whose levels usually correlate with those of cellular hTfR1. (3) The PCSK7 SNP associated with circulating s-hTfR1 levels8 was also recently associated with liver cirrhosis in hemochromatosis patients mutated in HFE.36 Whether PC7 mutations affect this disease through their role on hTfR1 shedding remains to be verified.

Each PC has multiple substrates. Recently, we showed that PC7 knockout mice exhibit anxiolytic and novelty-seeking behaviors due to modulation of dopamine D4 receptors in brain, likely implicating cell-surface proteins.37 The discovery of hTfR1 as the first in vivo substrate of PC7 is highly valuable to progress in the identification of other transmembrane substrates. However, in view of the fact that human, but not mouse, TfR1 is shed by PC7, it will be important to compare the functions of this highly conserved enzyme in more than one species, in which the cognate PC7-substrates may have evolved differently.


We thank Marie-Claude Asselin, Ann Chamberland, and Josée Hamelin for excellent technical assistance. We also thank Dr. Robert Day (University of Sherbrooke) for generous supply of PACE4 knockout mice used to isolate primary hepatocytes and Dr. Annelise Chapalain (INRS – Institut Armand-Frappier) for drawing the picture shown in Suporting Fig. 8. We thank all the members of the Seidah laboratory for helpful discussions and Brigitte Mary for editorial assistance.