Hepatocyte growth factor is a preferred in vitro substrate for human hepsin, a membrane-anchored serine protease implicated in prostate and ovarian cancers

Abstract

Hepsin is a membrane-anchored, trypsin-like serine protease with prominent expression in the human liver and tumours of the prostate and ovaries. To better understand the biological functions of hepsin, we identified macromolecular substrates employing a tetrapeptide PS-SCL (positional scanning-synthetic combinatorial library) screen that rapidly determines the P1-P4 substrate specificity. Hepsin exhibited strong preference at the P1 position for arginine over lysine, and favoured threonine, leucine or asparagine at the P2, glutamine or lysine at the P3, and proline or lysine at the P4 position. The relative activity of hepsin toward individual AMC (7-amino-4-methylcoumarin)-tetrapeptides was generally consistent with the overall peptide profiling results derived from the PC-SCL screen. The most active tetrapeptide substrate Ac (acetyl)-KQLR-AMC matched with the activation cleavage site of the hepatocyte growth factor precursor sc-HGF (single-chain HGF), KQLR downward arrowVVNG (where downward arrow denotes the cleavage site), as identified by a database analysis of trypsin-like precursors. X-ray crystallographic studies with KQLR chloromethylketone showed that the KQLR peptide fits well into the substrate-binding cleft of hepsin. This hepsin-processed HGF induced c-Met receptor tyrosine phosphorylation in SKOV-3 ovarian cancer cells, indicating that the hepsin-cleaved HGF is biologically active. Activation cleavage site mutants of sc-HGF with predicted non-preferred sequences, DPGR downward arrowVVNG or KQLQ downward arrowVVNG, were not processed, illustrating that the P4-P1 residues can be important determinants for substrate specificity. In addition to finding macromolecular hepsin substrates, the extracellular inhibitors of the HGF activator, HAI-1 and HAI-2, were potent inhibitors of hepsin activity (IC50 4+/-0.2 nM and 12+/-0.5 nM respectively). Together, our findings suggest that the HGF precursor is a potential in vivo substrate for hepsin in tumours, where hepsin expression is dysregulated and may influence tumorigenesis through inappropriate activation and/or regulation of HGF receptor (c-Met) functions.

Figure 1. P1–P4 substrate specificity of recombinant human wild-type hepsin as determined using a PS-SCL screen

A PS-SCL screen was used to determine the amino acid specificity at each of the S1–S4 subsites. The y-axis shows pM ACC-fluorophore released/s (conversion factor RFU to pM: 0.000723). The primary (P1)-specificity was determined with a P1-complete diverse library containing 137180 different peptides. The P2–P4 specificities were determined with P1/R-fixed libraries, where P2, P3 or P4 was randomized in a defined manner. Each P1/R-P2, -P3, or -P4 library contained 6859 different peptides. A total of 157757 peptide combinations were screened. The one-letter code is used to denote the 20 natural amino acids, minus cysteine and plus norleucine (n; 2-aminohexanoic acid) on the x-axis. Hepsin preferred arginine over lysine in the P1 position and exhibited moderate specificity in the P2, P3 and P4 positions. Favoured amino acids are indicated by * after setting an arbitrary threshold at 2000 pM ACC-fluorophore released per s (depicted as a horizontal line).

Figure 2. Relative amidolytic activity of hepsin toward selected AMC-coupled peptides

Peptides with the general structure Ac-P4-P3-P2-P1-AMC were tested in hepsin activity assays. The linear rates of fluorescence increase were expressed as percentage (%) activities when compared to the activity obtained with the Ac-KQLR-AMC peptide (set to 100%). The peptides were assayed for hydrolysis by hepsin at least twice in triplicate.

Figure 3. X-ray crystal structure of hepsin with KQLR-cmk

(A) A ribbon model representing the three-dimensional structure of hepsin. This structure is consistent with the report of Somoza et al. [30]. In its mature form, hepsin is a two-domain protein with a 109-amino-acid SRCR (extracellular scavenger receptor cysteine-rich domain) coupled via a single disulphide bond to a 255 amino-acid serine protease domain. The protein is predicted to be anchored to the membrane through a stretch of 27 amino acids at its N-terminus that potentially forms an α-helix (depicted as a broken line). The protease active site is formed in a cleft between the two six-stranded β-barrels. (B) A surface representation of the peptide-binding cleft of hepsin in complex with the KQLR-cmk peptide inhibitor. Highlighted in purple are hepsin residues Glu²⁵², Glu²⁵³, Tyr³⁰¹ and Trp³⁷⁷, selected amino acids of the active-site cleft. The P1-arginine is buried in the S1 pocket of hepsin, the P2-leucine lies in a dimple present in the S2 pocket, and the P3-glutamine interacts with the S3 pocket through a hydrogen bond with Tyr³⁰¹. The P4 lysine, which shows weaker electron density, is modelled into the S4 pocket above Trp³⁷⁷. The figure was generated with PyMOL (DeLano Scientific LLC, South San Francisco, CA, U.S.A.).

Figure 4. Hepsin digests of wild-type sc-HGF precursor (A, B), activation site mutant HGF precursors (B) and plasminogen (C)

Precursors were incubated for 2 h at 37 °C in the absence (−) (lanes 1, 3, 4, 5, 9) or in the presence (+) (lanes 2, 6, 7, 8) of hepsin. The cleavage products were analysed by SDS/PAGE under reducing conditions. (A) Wild-type HGF precursor (lane 1; 94–85 kDa; activation domain sequence KQLR↓VVNG) is converted into α/β-heterodimeric HGF (54 kDa α-chain; 26 kDa β-chain) by hepsin (lane 2). The N-terminal amino acid sequence of the β-chain fragment was determined as H₂N-VVNG, which indicates that the protein was specifically processed through cleavage between the P1 and P1′ residues in the activation domain. (B) Immunoblot (IB) analysis using an anti-HGF α-chain-specific antibody: two activation domain mutants of the HGF precursor (mutant 1, KQLQ↓VVNG; lanes 4 and 7, mutant 2, DPGR↓VVNG, lanes 5 and 8) are not processed by hepsin. In a parallel digest, wild-type HGF precursor was converted into α/β-heterodimeric HGF (lanes 3 and 6). (C) Plasminogen (lane 9; activation domain sequence CPGR-VVGG) is not converted to plasmin when incubated with hepsin (lane 10). The non-cleavage of plasminogen or the sc-HGF mutant precursors shows that the P4–P1 residues are specificity determinants for HGF processing by hepsin.

Figure 5. Hepsin-processed wild-type HGF stimulates c-Met receptor tyrosine auto-phosphorylation in SKOV-3 ovarian cancer cells

(A) SKOV-3 cells were treated for 15 min at 37 °C with (lane 1) 10% FBS-containing medium, or in serum-free medium with heterodimeric recombinant HGF (lane 2), sc-HGF (lane 3), and hepsin-processed HGF (lane 4). (B) SKOV-3 cells in serum-free medium (lane 5) were treated with heterodimeric recombinant HGF (lane 6), heterodimeric HGF plus anti-HGF neutralizing antibody (lane 7), hepsin-processed HGF (lane 8), and (lane 9) hepsin-processed HGF plus anti-HGF neutralizing antibody (lane 9). Immunoblot (IB) analysis was performed using anti-phospho-tyrosine c-Met antibody or anti-c-Met antibody. The arrow points to the 145 kDa β-chain of c-Met receptor (active form) under reducing conditions.