Pericellular activation of hepatocyte growth factor by the transmembrane serine proteases matriptase and hepsin, but not by the membrane-associated protease uPA.

Abstract

HGF (hepatocyte growth factor) is a pleiotropic cytokine homologous to the serine protease zymogen plasminogen that requires canonical proteolytic cleavage to gain functional activity. The activating proteases are key components of its regulation, but controversy surrounds their identity. Using quantitative analysis we found no evidence for activation by uPA (urokinase plasminogen activator), despite reports that this is a principal activator of pro-HGF. This was unaffected by a wide range of experimental conditions, including the use of various molecular forms of both HGF and uPA, and the presence of uPAR (uPA receptor) or heparin. In contrast the catalytic domains of the TTSPs (type-II transmembrane serine proteases) matriptase and hepsin were highly efficient activators (50% activation at 0.1 and 3.4 nM respectively), at least four orders of magnitude more efficient than uPA. PS-SCL (positional-scanning synthetic combinatorial peptide libraries) were used to identify consensus sequences for the TTSPs, which in the case of hepsin corresponded to the pro-HGF activation sequence, demonstrating a high specificity for this reaction. Both TTSPs were also found to be efficient activators at the cell surface. Activation of pro-HGF by PC3 prostate carcinoma cells was abolished by both protease inhibition and matriptase-targeting siRNA (small interfering RNA), and scattering of MDCK (Madin-Darby canine kidney) cells in the presence of pro-HGF was abolished by inhibition of matriptase. Hepsin-transfected HEK (human embryonic kidney)-293 cells also activated pro-HGF. These observations demonstrate that, in contrast with the uPA/uPAR system, the TTSPs matriptase and hepsin are direct pericellular activators of pro-HGF, and that together these proteins may form a pathway contributing to their involvement in pathological situations, including cancer.

Full text

Page 1

Pericellular activation of hepatocyte growth factor by the transmembrane serine
proteases matriptase and hepsin, but not by the membrane-associated protease uPA

Kate A. Owen*, Deyi Qiu*, Juliano Alves†, Andrew M. Schumacher†, Lynette M. Kilpatrick*,
Jun Li†, Jennifer L. Harris† and Vincent Ellis*

*Biomedical Research Centre, School of Biological Sciences, University of East Anglia,
Norwich NR4 7TJ, United Kingdom

†Genomics Institute of the Novartis Research Foundation, 10675 John Jay Hopkins Drive, San
Diego, CA 92121, USA


ABSTRACT

Hepatocyte growth factor (HGF) is a pleiotropic cytokine homologous to the serine protease
zymogen plasminogen that requires canonical proteolytic cleavage to gain functional activity.
The activating proteases are key components of its regulation, but controversy surrounds their
identity. Using quantitative analysis we found no evidence for activation by the urokinase-type
plasminogen (uPA), despite reports that this is a principal activator of pro-HGF. This was
unaffected by a wide range of experimental conditions, including the use of various molecular
forms of both HGF and uPA, and the presence of uPA receptor (uPAR) or heparin. By contrast
the catalytic domains of the type-II transmembrane serine proteases (TTSPs) matriptase and
hepsin were highly efficient activators (50% activation at 0.1 and 3.4 nM, respectively), at least
4-orders of magnitude more efficient than uPA. Positional scanning-synthetic combinatorial
peptide libraries identified consensus sequences for the TTSPs, which in the case of hepsin
corresponded to the pro-HGF activation sequence, demonstrating a high specificity for this
reaction. Both TTSPs were also found to be efficient activators at the cell surface. Activation
of pro-HGF by PC-3 prostate carcinoma cells was abolished by both protease inhibition and
matriptase-targeting siRNA, and scattering of MDCK in the presence of pro-HGF was
abolished by inhibition of matriptase. Hepsin-transfected 293 cells also activated pro-HGF.
These observations demonstrate that, in contrast to the uPA/uPAR system, the TTSPs
matriptase and hepsin are direct pericellular activators of pro-HGF, and that together these
proteins may form a pathway contributing to their involvement in pathological situations,
including cancer.


Keywords: Hepatocyte growth factor (HGF), type-II transmembrane serine proteases (TTSPs),
substrate specificity, proteolytic processing, pericellular proteolysis, plasminogen activation.


Short title: Activation of pro-HGF by pericellular serine proteases


Abbreviations: HGF, hepatocyte growth factor; HGFA, HGF activator; uPA, urokinase
plasminogen activator; uPAR, uPA receptor; TTSP, type-II transmembrane serine protease;
MDCK, Madin-Darby canine kidney cells
Biochemical Journal Immediate Publication. Published on 17 Dec 2009 as manuscript BJ20091448
THIS IS NOT THE VERSION OF RECORD - see doi:10.1042/BJ20091448
Accepted Manuscript
Licenced copy. Copying is not permitted, except with prior permission and as allowed by law.
© 2009 The Authors Journal compilation © 2009 Portland Press Limited

Page 2

INTRODUCTION
Hepatocyte growth factor (HGF), also known as scatter factor, is a polypeptide growth factor
produced by mesenchymal cells that exerts pleiotropic effects in multiple tissues (reviewed in
[1]). It is an important paracrine mediator of epithelial-mesenchymal cell interactions, affecting
many cellular processes, including proliferation, differentiation, motility, invasion and
branching morphogenesis. As well as being essential for development [2;3], HGF is also
implicated in a variety of physiological and pathological processes including wound healing,
cancer and cardiovascular disease. These effects are all mediated by the binding of HGF to a
single receptor, the proto-oncogene receptor tyrosine kinase c-Met.
Accepted Manuscript
HGF is closely related to the serine protease zymogen plasminogen, comprising an N-terminal
PAN/apple domain, four kringle domains (compared with five in plasminogen) and a C-
terminal serine protease domain. However, the latter is enzymatically inactive due in part to
mutations to two of the three residues (His57 and Ser195)1 comprising the serine protease
catalytic triad. The primary binding determinants for the interaction of HGF with c-Met are
contained within the N-terminal and kringle domains of HGF [4]. Although mediating high-
affinity binding, these interactions are not sufficient for the activation of c-Met.
A key characteristic that HGF retains from its serine protease ancestry is the absolute
requirement for limited proteolysis to acquire biological activity. Similarly to plasminogen and
other serine protease zymogens, HGF is secreted as a single-chain precursor form and this pro-
HGF form is devoid of signalling activity. Pro-HGF is activated to a disulphide-bridged, two-
chain molecule by proteolytic cleavage at the canonical Arg15-Val16 bond in the serine protease
domain. In the serine proteases, this cleavage results in conformational rearrangements in the
so-called “activation domain” and maturation of the active site. Structural studies have
demonstrated that HGF undergoes corresponding conformational changes on proteolytic
cleavage and that the rearranged activation domain constitutes an additional binding site for c-
Met [5;6]. The binding of HGF to c-Met is also thought to involve a large-scale interdomain
reorganisation [6], also analogous to that observed in plasminogen [7]. Together these
conformational changes, caused by a single proteolytic cleavage, enable HGF/c-Met signalling.
Proteolytic activation of pro-HGF clearly plays an essential role in the function of the HGF/c-
Met signalling pathway, however, the proteases responsible for this key regulatory step and
how this proteolysis is regulated are not well understood. The only protease unequivocally
demonstrated to be an authentic activator of pro-HGF is the serine protease HGFA. This is an
efficient activator of HGF in vitro [8], and has been shown to contribute to the activation of
pro-HGF in vivo [9]. HGFA is activated by thrombin and, consistent with this, it has a role
primarily at sites of tissue injury [10]. HGFA-null mice develop normally, in contrast to c-Met
or HGF-null mice, but have a partial impairment of tissue repair [9].
Other serine proteases have also been demonstrated to activate pro-HGF in vitro, several of
which are pericellular proteases. These include both integral membrane proteins and proteases
associated with membrane receptors or binding sites, which have key roles in regulating cell
behaviour [11] and that may act to regulate the activity of HGF on the surface of target cells,
i.e. those expressing c-Met. Of particular interest in this respect is the plasminogen activator
uPA, which has been proposed to be an important activator of HGF [12-15]. These studies
were initially prompted by the homology of HGF and plasminogen, and would provide an
attractive mechanism for the pericellular activation of HGF, as the activity of uPA is promoted
by association with its specific GPI-anchored receptor uPAR on the surface of many cell types
[16]. However, other studies have found no evidence for activation of pro-HGF by uPA in

1 Numbering used for the serine protease domain follows the convention of numbering according to the sequence
of chymotrypsinogen

2
Biochemical Journal Immediate Publication. Published on 17 Dec 2009 as manuscript BJ20091448
THIS IS NOT THE VERSION OF RECORD - see doi:10.1042/BJ20091448
Licenced copy. Copying is not permitted, except with prior permission and as allowed by law.
© 2009 The Authors Journal compilation © 2009 Portland Press Limited

Page 3

purified systems [8;17]. The activity of plasmin itself has also been implicated in the activation
of pro-HGF [18;19]. Soluble forms of two type-II transmembrane serine proteases (TTSPs),
matriptase [20] and hepsin [21;22], have been shown to activate pro-HGF in purified systems,
but their contribution in the pericellular environment is unclear.
Accepted Manuscript
Here we assess the contribution of putative proteolytic activators of pro-HGF by quantitative
comparison of their activities in solution, their P4-P1 substrate specificities and their activities
in the pericellular environment. We find no evidence for activation of pro-HGF by uPA the
plasminogen activator uPA under a wide variety of conditions, making it unlikely that it has a
biologically relevant role. By contrast, the TTSPs matriptase and hepsin were found to be
efficient activators, both in solution and, more importantly, when expressed as transmembrane
proteases. Together these observations demonstrate that both matriptase and hepsin are
effective proteolytic activators of pro-HGF at the cell surface, and furthermore that hepsin has
specificity characteristics suggesting it may be a highly selective activator of pro-HGF.
MATERIALS AND METHODS
Proteins and reagents
Human urinary uPA was from Serono SpA (Rome, Italy), pro-uPA from Abbott (Chicago, IL),
tPA from Boerhinger-Ingelheim GmbH (Ingelheim, Germany), factor XIIa and plasmin from
Enzyme Research Laboratories (Swansea, UK). Human HGFA was obtained from R&D
Systems (Abingdon, UK) and activated using immobilised thrombin (Calbiochem,
Nottingham, UK). Recombinant human matriptase protease domain was expressed in E .coli
and yeast as previously described [23]. Soluble uPAR expressed in Drosophila S2 cells was a
gift from Dr. Michael Ploug (Finsen Laboratory, Copenhagen). Pro-HGF expressed in
Saccharomyces cerevisiae was a gift from Dr. George Van Woude (Van Andel Research
Institute, Grand Rapids, MI). The matriptase inhibitor CJ-730 (compound #8 in [24]) was a gift
of Dr. Kerstin Uhland (Curacyte AG, Munich, Germany). The anti-matriptase monoclonal
antibody M32 was a kind gift of Dr. Chen-Yong Lin (Georgetown University Medical Centre).
Expression and purification of pro-HGF
HGF cDNA was PCR amplified from a random hexamer primed cDNA library, constructed
from human fetal lung MRC-5 cells. PCR products were ligated into pGEM-T Easy (Promega),
transformed into E. coli DH5α (Invitrogen, Paisley, UK) and subcloned into the expression
vector pMT-V5HisB (Invitrogen). The resulting plasmid pMT-HGF-V5His was cotransfected
with pCoHygro into Drosophila Schneider S2 cells using CellFectin in serum-free Drosophila
Expression Medium (all reagents from Invitrogen). After overnight incubation, cells were
cultured for two days in complete medium at 27°C in atmospheric air then selected in 300-500
µg/ml hygromycin B for 5 weeks. Cells stably expressing HGF were adapted to serum-free
medium and expression induced with 0.5 mM CuSO4. Conditioned medium was collected after
7 days. Recombinant pro-HGF was purified from cleared S2-conditioned medium using metal
chelate and heparin affinity chromatography. Conditioned medium was incubated overnight
with Ni-NTA agarose (Qiagen) at 4°C, loaded into a column and washed with 50 mM sodium
phosphate pH 8.0, 0.3 M NaCl, 20 mM imidazole. HGF was eluted in the same buffer
containing 0.25 M imidazole. The eluted protein was incubated with heparin-agarose (Sigma)
in 10 mM Tris-HCl pH 8.0 overnight at 4°C, loaded into a column and washed with 10 mM
Tris-HCl pH 8.0, 0.3 M NaCl. Bound HGF was eluted in 10 mM Tris-HCl pH8.0, 1.5 M NaCl.
Expression and purification of soluble hepsin
Human hepsin was subcloned without the transmembrane domain, spanning residues 47-417.
The honeybee melittin secretion signal sequence was appended to the N-terminus by PCR, and
a 6X-His tag was appended to the C-terminus. All PCR products were inserted into pFastBac1

3
Biochemical Journal Immediate Publication. Published on 17 Dec 2009 as manuscript BJ20091448
THIS IS NOT THE VERSION OF RECORD - see doi:10.1042/BJ20091448
Licenced copy. Copying is not permitted, except with prior permission and as allowed by law.
© 2009 The Authors Journal compilation © 2009 Portland Press Limited

Page 4

vector (Invitrogen) via EcoRI and NotI sites and sequenced. The plasmids were then used to
produce recombinant baculovirus following the manufacturer’s protocol (Invitrogen). Sf9 cells
were infected at a MOI of 5-10 and the activity of hepsin in the supernatant monitored by
hydrolysis of the fluorogenic peptide Ac-KRLR-ACC. When maximal activity was achieved,
the supernatant was collected, adjusted to pH 5.0 with 4.0 M NaAcetate and cleared by
centrifugation. Filtered media was loaded onto a POROS HS column on an Akta FPLC (GE
Healthcare) at 17 ml/min and washed with 75 mM NaAcetate pH 5.0 until baseline recovered.
Bound proteins were eluted with 50 mM Tris pH 8.0, 1.0 M NaCl and dialyzed against 50 mM
Tris, 20 mM NaCl overnight. Dialyzed sample was loaded onto a benzamidine column at 1.5
ml/min, washed with 50 mM Tris pH 8.0, 0.2 M NaCl and eluted with 0.2 M Glycine pH 3.0.
Eluted fractions were immediately neutralized with 2.0 M Tris, pH 8.0. The eluate was
concentrated and further purified by Sephadex 200 chromatography (GE Healthcare) in 20 mM
Tris pH 8.0, 0.15 M NaCl. Active hepsin fractions were pooled and concentrated.
Active Site Titration
The molar concentration of active sites for each protease was determined by active site titration
using 4-methylumbelliferyl p-guanidinobenzoate (Sigma). Proteases were dissolved in 50 mM
Hepes pH 7.4, 0.1 M NaCl, 0.01% Tween 80 and fluorescence recorded at 25°C in a Perkin-
Elmer LS50B luminescence spectrometer.
Proteolytic activation of pro-HGF
Recombinant single-chain pro-HGF (10 nM) was incubated with varying concentrations of the
stated protease in 10 mM sodium phosphate pH 8.0, 0.12 M NaCl, 5 mM EDTA, 0.01% Tween
80 for 1hr at 37°C. In some experiments the reactions were also carried out in 50 mM Tris-
HCl, pH 8.8, 38 mM NaCl, 0.01% Tween 80. Reactions were terminated by boiling in SDS-
PAGE sample buffer containing 0.1 M dithiothreitol. The samples were run on 10% SDS-
PAGE, transferred to PVDF membranes(Bio-Rad, Hemel Hempstead, UK) and the activation
status of HGF determined by Western blotting using an anti-V5 monoclonal antibody
(Invitrogen) and HRP-conjugated rabbit anti-mouse IgG (Dako, Glostrup, Denmark). The blots
were incubated with ECL Plus (GE Healthcare) and visualised either by exposure to Hyperfilm
ECL (GE Healthcare) or chemiluminescence quantified using a Storm PhosphorImager® (GE
Healthcare) and ImageQuant® software (Molecular Dynamics, Sunnyvale, CA).
Quantitative real-time PCR
The expression of matriptase, hepsin, uPA and uPAR was determined in prostate-derived cell
lines using reverse transcribed total RNA and specific primer/probe sets in the 7500 Fast Real-
Time PCR System (Applied Biosystems, Warrington, UK), as previously described [25].
Cellular activation of pro-HGF
MDCK, PC3 or hepsin-transfected 293 cells were plated in 96 well plates at 20,000 cells/well
and cultured overnight. Cells were washed with PBS before incubation with pro-HGF in a total
of 100 µl medium for up to 8 hrs. Conditioned medium was removed, subjected to SDS-PAGE
and analysed for activation of pro-HGF by Western blotting.
Ki determinations and combinatorial peptide library analysis
Inhibition constants of the inhibitors CJ-730 and amiloride were determined using the substrate
Ac-Pro-Arg-Leu-Arg-AMC (for matriptase, hepsin and HGFA) and H-Glu-Gly-Arg-AMC (for
uPA). Substrate hydrolysis was monitored at 25°C in a SpectraMax Gemini microplate
spectrofluorometer (Molecular Devices). Ki values were determined from plots of Km/Vmax
against [I]. Positional scanning combinatorial peptide libraries were synthesized as previously
described [26] and profiled with recombinant hepsin and matriptase (expressed in yeast).

4
Biochemical Journal Immediate Publication. Published on 17 Dec 2009 as manuscript BJ20091448
THIS IS NOT THE VERSION OF RECORD - see doi:10.1042/BJ20091448
Accepted Manuscript
Licenced copy. Copying is not permitted, except with prior permission and as allowed by law.
© 2009 The Authors Journal compilation © 2009 Portland Press Limited

Page 5

siRNA knockdown of matriptase expression
PC3 cells were transfected with either of two matriptase-specific or universal control siRNA as
previously described [23]. Knockdown of matriptase expression was confirmed by Western
blotting using the antibody M32, and was greater than 90% compared to both control siRNA
and mock-transfected cells. Cells were analysed for pro-HGF activation 48 hours after
transfection.
Generation of cell lines stably expressing hepsin
Full-length hepsin cDNA (IMAGE clone #5228525) was ligated into pcDNA3.1(+)-HA to
generate the plasmid pcDNA3.1/HPN-HA and transfected into 293 cells using Fugene 6
(Roche) and cells selected for stable expression in 200 µg/ml Zeocin (Invitrogen). Clones were
tested for hepsin expression by Western blotting using an anti-HA antibody and TaqMan
quantitative real-time PCR using an ABI Prism® 7700 (Applied Biosystems). Two clones
stably expressing low and high levels of hepsin were selected (F10 and C6, respectively).
MDCK cell scatter assay
Madin-Darby canine kidney (MDCK-2) cells were seeded at a density of 103 cells per well and
left to adhere overnight. Cells were washed with serum-free medium and treated with 10 ng/ml
of either active or pro-HGF, in the presence or absence of CJ-730 (50 µM). After 30 hrs cells
were washed with PBS and fixed with ice cold methanol. Images were taken using a Zeiss
CCD inverted microscope using x10 magnification. Three independent experiments were
performed with all treatments in duplicate.
RESULTS
Expression of recombinant single-chain pro-HGF
The study of the proteolytic activation of HGF is greatly facilitated by the availability of
homogenous preparations of the single-chain form of the protein, a problem which has
hampered some previous studies. We expressed HGF in Drosophila S2 cells and under serum-
free conditions the recombinant protein was exclusively in the 92 kDa non-activated, single
chain form detected on SDS-PAGE with reduced samples (Figure 1). By contrast in serum-
containing medium there was significant activation of HGF, detected as the appearance of both
55 kDa N-terminal heavy chain and 34 kDa C-terminal light chain. In some of the experiments
shown subsequently, a small amount of activated HGF can be detected due to the high sample
loading which was necessary to ensure the detection of low levels of pro-HGF activation.
uPA is not an efficient activator of pro-HGF
To investigate the role of uPA in the activation of HGF, a fixed concentration of pro-HGF was
incubated with varying concentrations of uPA for 1-hr at 37°C and samples analysed by SDS-
PAGE and Western blotting. The latter was necessary as pro-HGF was used at a relatively low
concentration (10 nM) in these experiments, so as to be likely below the Km for the reaction
and therefore increase the fractional amount of substrate hydrolysed. However, no activation of
HGF was observed with concentrations of uPA up to 50 nM (Figure 2A), with no reduction in
the intensity of the 92 kDa pro-HGF band nor an increase in the 34 kDa HGF light-chain band.
Increasing incubation times up to 5-hr also failed to demonstrate any activation of pro-HGF
(data not shown). Subsequently the concentration of uPA was increased up to 3 µM (Figure
2B). Even under these conditions no activation of pro-HGF was apparent, with no increase in
the intensity of the 34 kDa light-chain band. However, a reduction in the intensity of both this
and the pro-HGF band was observed, which was most likely due to non-specific proteolytic
cleavage of the C-terminal epitope tag at this extremely high concentration of uPA. Consistent
with this interpretation, non-reduced samples also showed a decrease in intensity of the single
band representing both molecular forms of HGF (data not shown).

5
Biochemical Journal Immediate Publication. Published on 17 Dec 2009 as manuscript BJ20091448
THIS IS NOT THE VERSION OF RECORD - see doi:10.1042/BJ20091448
Accepted Manuscript
Licenced copy. Copying is not permitted, except with prior permission and as allowed by law.
© 2009 The Authors Journal compilation © 2009 Portland Press Limited