Alphafold2-Assisted Structure-Function Analysis of Human Erythroferrone Relevant for Its Iron-Regulatory Function

Erythroferrone (ERFE), an erythrokine that helps mobilize iron for developing erythrocytes, acts by decreasing the production of hepcidin, through sequestration of bone morphogenetic proteins (BMPs) that bind to BMP receptors to activate hepcidin transcription. We used Alphafold2 (AF2) to model the conformation of human erythroferrone monomers and various multimers, with or without interactions with BMPs, with emphasis on the principal ligand, the BMP2/6 heterodimer. We then tested the models by targeted mutations to disrupt key interactions suggested by the models, and assessed the effect of the mutations on bioactivity and BMP binding. ERFE is a member of the C1q/TNFa family of proteins known for containing N-terminal unstructured regions and C-terminal TNFa-like globular heads that promote the formation of a spectrum of biologically significant multimers. After cleavage by a prohormone convertase, the ERFE molecule consists of an unstructured N-terminal segment residues 43-185 and a globular TNF-like head 186-354. We found that a segment of the N-terminal region of the molecule-segment 43-148 of the 354 amino acid protein-was sufficient to bind BMPs and suppress hepcidin; the larger C-terminal region does neither. For bioassays, we prepared ERFE or its mutants in HEK293 mammalian cells and tested these without purification, so as not to disrupt their mixture of multimeric conformations. The Hep3B human liver cell line was used for bioactivity assays with the readout of hepcidin mRNA concentrations, measured by qRT-PCR. To avoid the confounding effects of ERFE multimerization driven by the TNFa-like globular regions on avidity, we purified untagged ERFE N-terminal (43-148) segments or their mutants from bacteria, and measured the binding between these and BMPs using surface plasmon resonance. These forms lacked TNFa-like heads and, unlike full length ERFE, did not form multimers in nondenaturing PAGE.

We identified and explored several structural elements within the ERFE N-terminus. All models of BMP interaction with ERFE predicted the binding of a tryptophan-containing hydrophobic helix of ERFE to a deep groove formed at the interface of all BMP dimers (BMP2/6 in cyan/green, ERFE in magenta; interaction in callout box, left panel). The highly evolutionarily conserved hydrophobic region (residues 81-86) was required for bioactivity and for tight binding to BMPs. Deletion of the hydrophobic segment or W82A substitution of the strictly conserved tryptophan ablated bioactivity and greatly decreased the binding of ERFE to BMPs. In nearly all models, a positively-charged cationic segment and the collagen-like segment made no contact with BMPs. However, the cationic segment was required for activity (right panel) although its deletion had only minor effect on binding to BMP. The specific arrangement of arginines and lysines in this segment did not matter for bioactivity or binding. We showed that the segment was essential for heparin-binding and propose that it may serve to concentrate ERFE at the site of its activity. Deletion of the collagen segment or the more distal helical segment (green) impaired IC ₅₀ much less (50-100fold) but also substantially disrupted monomeric binding. The collagen segment forms a triple helix in trimeric models of ERFE, but the helix is disrupted by BMP binding.

Secondary binding contacts between ERFE and BMPs in AF2 varied depending on the oligomeric state of erythroferrone. In monomers and dimers, the helical segment made contact with the contralateral wing of the BMP dimer. Remarkably, both primary and secondary binding interactions of BMPs were structurally similar to the BMP interactions with their receptors or known antagonists.

Nondenaturing PAGE suggested that the predominant form of ERFE was a hexamer. The AF2 structure of hexameric ERFE (magenta) with three molecules of BMP2/6 (cyan-green) showed that each BMP heterodimer bound two ERFE proximal helical segments including 81-86, within each of its two BMP2/6 interfacial grooves (left panel) but unlike in monomers and dimers, the secondary interaction did not involve the distal helical segment but the globular heads (red trimers on each side).

Despite their reductionist nature, these studies provide useful structural predictions that should allow better pharmacologic targeting of this important erythroid regulator of iron homeostasis.