High-resolution analysis and functional mapping of cleavage sites and substrate proteins of furin in the human proteome

Abstract

Background: There is a growing appreciation of the role of proteolytic processes in human health and disease, but tools for analysis of such processes on a proteome-wide scale are limited. Furin is a ubiquitous proprotein convertase that cleaves after basic residues and transforms secretory proproteins into biologically active proteins. Despite this important role, many furin substrates remain unknown in the human proteome.

Methodology/principal findings: We devised an approach for proteinase target identification that combines an in silico discovery pipeline with highly multiplexed proteinase activity assays. We performed in silico analysis of the human proteome and identified over 1,050 secretory proteins as potential furin substrates. We then used a multiplexed protease assay to validate these tentative targets. The assay was carried out on over 3,260 overlapping peptides designed to represent P7-P1' and P4-P4' positions of furin cleavage sites in the candidate proteins. The obtained results greatly increased our knowledge of the unique cleavage preferences of furin, revealed the importance of both short-range (P4-P1) and long-range (P7-P6) interactions in defining furin cleavage specificity, demonstrated that the R-X-R/K/X-R ↓ motif alone is insufficient for predicting furin proteolysis of the substrate, and identified ≈ 490 potential protein substrates of furin in the human proteome.

Conclusions/significance: The assignment of these substrates to cellular pathways suggests an important role of furin in development, including axonal guidance, cardiogenesis, and maintenance of stem cell pluripotency. The novel approach proposed in this study can be readily applied to other proteinases.

Figure 1. A flowchart representing the stepwise approach used to identify furin substrates in the human proteome.

Figure 2. Frequency plot of the cleavage sequence of furin in an IceLogo format.

The height of a character is proportional to the frequency of the amino acid residue at the individual position of the cleaved peptide and is normalized for the amino acid encoded in the entire human genome according to RefSeq. Because there was a five residue overlap between the P7-P1’ and P4-P4’ peptides from the individual cleavage site, the sequence logos represent the combined P7-P4’ peptide sequence. A. Peptide sequences with the Z-score over 3.5 for both P4-P4’ and P7-P1’ peptides (932 substrates). B. Peptide sequences with the Z-score for the P4-P4’ and P7-P1’ peptides over and below 3.5, respectively (64 substrates). C. Peptide sequences with the Z-score for the P4-P4’ and P7-P1’ peptides below and over 3.5, respectively (439 substrates). D. Peptide sequences for poor or no substrates with Z-score for the P4-P4’ and P7-P1’ peptides below 1.0 (391 sequences).

Figure 3. Structural modeling of furin with a protein substrate.

A. The structure of furin with the SSNSRKRR↓S modeled substrate. The structure of furin-DEC complex (PDB 1P8J) was used as a template for modeling the protein substrate (shown as sticks) in the furin structure. The molecular surface of furin is colored according to the electrostatic potential (red, blue and white are negative, positive and neutral electrostatic potential values, respectively). Close-up on the left shows the S6 Glu230 and S5 Glu257 (yellow), which are critical for the interactions with the P6 and P5 substrate sub-sites. Close-ups (B and C) show in a more detail the P5-P4 and P2-P1’ residue positions in the furin active site.

Figure 4. Furin processing of integrin αV.

The original glioma U251 cells, breast carcinoma MCF-7 cells with the enforced expression of the β3 integrin subunit (MCF7-β3), colon carcinoma LoVo cells with and without the enforced expression of furin and CHO cells with the enforced expression of the integrin αV precursor (pro-αV) or the furin-resistant pro-αV mutant (pro-αV-A889A890) were cell-surface biotinylated and then lysed. The lysates were precipitated using the αVβ3 antibody LM609 and analyzed by Western blotting with Extravidin-horseradish peroxidase. Where indicated, cells were co-incubated for 16–18 h prior to biotinylation with a synthetic furin inhibitor (DEC; 50 μM).

Figure 5. Functional analysis of predicted furin human substrates.

Ingenuity Pathway Analysis was used to identify the most significant developmental functions (A), molecular and cellular functions (B) and canonical pathways (C) associated with the predicted furin protein substrates. Right-tailed Fisher exact test was used to calculate a p-value. Only pathways with P<0.05 are shown.

Figure 6. Examples of canonical pathways which are likely to be regulated by furin proteolysis.

A. Pluripotency in human embryonic stem cells; B. the bone morphogenic protein (BMP) pathway in cardiomyocyte differentiation; C. Notch signaling pathway. Furin substrates are pink.