Báo cáo khoa học: Hepatocyte growth factor activator (HGFA): molecular structure and interactions with HGFA inhibitor-1 (HAI-1) doc

MINIREVIEW
Hepatocyte growth factor activator (HGFA): molecular
structure and interactions with HGFA inhibitor-1 (HAI-1)
Charles Eigenbrot
1,2
, Rajkumar Ganesan
3
and Daniel Kirchhofer
3
1 Department of Structural Biology, Genentech, Inc., South San Francisco, CA, USA
2 Department of Antibody Engineering, Genentech, Inc., South San Francisco, CA, USA
3 Department of Protein Engineering, Genentech, Inc., South San Francisco, CA, USA
Introduction
Hepatocyte growth factor activator (HGFA) is a
trypsin-like serine protease belonging to Clan PA,
Family S1 (MEROPS data base, ger.
ac.uk/). Full-length HGFA (96 kDa) has the same
domain architecture as coagulation factor XII: an
N-terminal fibronectin type II domain, an epidermal
growth factor (EGF)-like domain, a fibronectin type I
domain, a second EGF-like domain, a kringle domain,
and a C-terminal protease domain (Fig. 1A). The
HGFA protease domain amino acid sequence has the
Keywords
catalysis; hepatocyte growth factor; Kunitz
domain; serine protease; structure
Correspondence
D. Kirchhofer, Genentech, Inc., 1 DNA Way,
South San Francisco, CA 94080, USA
Fax: +1 (650) 225-3734
Tel: +1 (650) 225-2134

E-mail:
(Received 13 November 2009, revised
19 January 2010, accepted 8 February
2010)
doi:10.1111/j.1742-4658.2010.07638.x
The trypsin-like serine protease hepatocyte growth factor activator
(HGFA) undergoes proteolytic activation during blood coagulation, result-
ing in a 34 kDa ‘short form’, consisting mainly of the protease domain.
The crystal structures of the recombinantly expressed HGFA ‘short form’
discussed herein have provided molecular insights into its interaction with
inhibitors and substrates, as well as the regulation of catalytic activity. The
HGFA structures revealed enzymatically competent and noncompetent
forms associated with the conformational states of two substrate specific-
ity-determining loops, the 220-loop and 99-loop. The implied dynamic
behavior of these loops, which are intimately involved in substrate interac-
tion, has precedents in other members of the S1 family of serine proteases,
and may be associated with specific mechanisms of enzyme regulation.
Furthermore, HGFA activity is strongly inhibited by HGFA inhibitor-1, a
membrane-spanning multidomain inhibitor containing two Kunitz
domains, of which only the N-terminal Kunitz domain-1 (KD1) inhibits
enzymatic activity. In the structure of the KD1–HGFA complex, the
inhibitor interacts with the active site region by making contacts with all
substrate specificity-determining loops and by occupying subsites S1, S2
and S4 in a substrate-like manner. In fact, the side chains of KD1 residues
occupying these sites are virtually superimposable on the P1, P2 and P4
residues of the pro-hepatocyte growth factor-derived substrate mimic
Lys-Gln-Leu-Arg chloromethyl ketone bound to HGFA. These structures
also allow us to rationalize the apparently narrow substrate specificity of
HGFA, which is limited to the two known macromolecular substrates
pro-hepatocyte growth factor and pro-macrophage-stimulating protein.

Abbreviations
EGF, epidermal growth factor; HAI, hepatocyte growth factor activator inhibitor; HGFA, hepatocyte growth factor activator; KD1, Kunitz
domain-1; KD2, Kunitz domain-2; LDL, low-density lipoprotein; PDB, Protein Data Bank; pro-HGF, pro-hepatocyte growth factor; pro-MSP,
pro-macrophage-stimulating protein; uPA, urokinase-type plasminogen activator.
FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS 2215
highest sequence identity with that of factor XII
(47%), tissue-type plasminogen activator (40%), uroki-
nase-type plasminogen activator (uPA) (39%), and
prostasin (38%). There is also a strong similarity at
the structural level, as indicated by the rmsd values
being lower than 1.0 A
˚
in a pairwise comparison of
the protease domain canonical crystal structures
(excluding loops): HGFA protease domain [Protein
Data Bank (PDB) 1YC0] versus tissue-type plasmino-
gen activator (PDB 1RTF, rmsd of 0.59 A
˚
), versus
uPA (PDB 1C5Y, rmsd of 0.58 A
˚
), and versus prosta-
sin (PDB 3DFL, rmsd of 0.67 A
˚
) (the structure of the
factor XII protease domain is not known). To date,
there are only two known macromolecular substrates
of HGFA, pro-hepatocyte growth factor (pro-HGF)
[1] and pro-macrophage-stimulating protein (pro-MSP)
[2], suggesting that HGFA has very limited substrate

specificity. However, this may be an underestimation
of the full complement of substrates, as no systematic
substrate profiling has been performed yet. The
AB
CD
Fig. 1. Conformational states of the HGFA active site region. (A) Cartoon of HGFA and HAI-1 domain architectures. HGFA contains a heavy
chain (A-chain) disulfide-linked to the protease domain (B-chain). The subdomains of the A-chain are: fibronectin type I and type II (FNI and
FNII), epidermal growth factor (EGF)-like and Kringle (Kr). Cleavage by thrombin (T) and plasma kallikrein (K) produces the serum form (‘short
form’) of HGFA, comprising the protease domain and a disulfide-linked 35 amino acid peptide (Val373–Arg407) from the A-chain, which was
used for crystallographic studies. HAI-1 is composed of a MANSC domain [25] followed by a structurally undefined region connecting to
KD1, an LDL receptor (LDLR)-like domain, KD2, a transmembrane (TM) domain, and cytoplasmic domain (Cyt). The splice variant HAI-1B has
an extra 16 amino acid stretch inserted (I) between KD1 and the LDLR-like domain. (B) The HGFA protease domain (beige, PDB 1YC0) with
colored substrate ⁄ inhibitor specificity-determining loops (chymotrypsinogen numbering, i.e. ‘38-loop’, and the corresponding Perona and
Craik [6] nomenclature, i.e. ‘Loop-A’) and substrate subsites (S1–S4). The catalytic triad Asp102–His57–Ser195 is indicated. (C) Conforma-
tional states of the 220-loop in HGFA (left panel) as compared with prostasin (right panel). Left panel: the ‘open’ (standard conformation;
PDB 1YC0) and ‘closed’ (nonstandard conformation; PDB 1YBW) HGFA forms are superimposed, with the two different 220-loop conforma-
tions shown in cyan and magenta, respectively. Side chains of the catalytic triad residues (Asp102, His57, and Ser195) are indicated (yellow
for ‘open’; magenta for ‘closed’), as is that of the 220-loop residues 215 and 216. The P1 Arg from the KQLR-cmk substrate mimic (see
Fig. 3A) is also added to indicate the steric clash with the ‘closed’ form 220-loop. Right panel: the ‘open’ (PDB 3DFL) and ‘closed’
(PDB 3DFJ) prostasin forms are superimposed, showing the two different 220-loop conformations. The color codes are the same as for
HGFA. The side chain of Asp217, which, in the ‘closed’ conformation, obstructs S1 access, is also indicated. (D) Conformational states of
the HGFA 99-loop. As compared with the competent (or standard) conformation (slate blue), the 99-loop of the Fab40-inhibited HGFA (brick
red: noncompetent) has shifted towards the substrate-binding cleft. The deleterious effects on catalysis derive from the repositioning of
P99a and S99, both of which shape the S2 subsite. Molecular images were produced using
PYMOL [38].
Molecular interactions of HGFA with inhibitor ⁄ substrate C. Eigenbrot et al.
2216 FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS
cleavage site of pro-HGF is KQLR-VVNG(491–498)
(P4–P4¢) and that of pro-MSP is SKLR-VVNG(487–
480) (P4–P4¢), indicating a preference for a P1 Arg.

The potential roles of these substrates in mediating the
proposed functions of HGFA in tissue regeneration
and tumor promotion are discussed in other sections
of this minireview series [3].
HGFA is secreted as a single-chain zymogen precur-
sor, and is activated by cleavage at the Arg407-Ile408
bond (e.g. by thrombin during blood coagulation),
resulting in the disulfide-linked two-chain form.
Additional cleavage by plasma kallikrein at the
Arg372-Val373 bond releases the 34 kDa HGFA ‘short
form’ present in serum, containing a 35-residue peptide
disulfide-linked to the protease domain. All structural
studies discussed herein were performed with the ‘short
form’ of HGFA recombinantly expressed in insect cells.
The seven available crystal structures of the HGFA
protease domain, either as apoenzyme or in complex
with an inhibitor, provide a basis for understanding
the known biochemical functions of this enzyme. Our
discussion is focused on the regulation of its catalytic
machinery, its interactions with HGFA inhibitor
(HAI)-1, and its substrate specificity. For mention of
specific amino acid positions, we use the chymotrypsin-
ogen numbering scheme to allow easy reference to the
large number of related proteins (for conversion
between native HGFA and chymotrypsinogen residue
numbers, see [4]), and we employ the nomenclature of
Schechter and Berger [5] in describing specific sites of
protease–substrate (or inhibitor) interactions. The
loops in and around the active site are named accord-
ing to their chymotrypsinogen numbering (for conver-

sion into the Perona and Craik [6] loop nomenclature,
see Fig. 1B).
The catalytically competent (standard)
active site conformation
The determined HGFA structures reveal three different
conformational states of the active site region: a cata-
lytically competent (standard) conformation, and two
nonstandard conformations. Herein, we use the con-
formation seen in the complex of HGFA with Kunitz
domain-1 (KD1) as the representative of the standard
conformation, which was also observed in complexes
with two antibody fragments (Fab58 and Fab75) [7].
The standard, i.e. the conventional, form of HGFA
(PDB 1YC0) displays features typical of the S1 family
of serine proteases, such as the double b-barrel arrange-
ment of the peptidase domain, a His57–Asp102–Ser195
catalytic triad, and distinct surface loops that determine
substrate and inhibitor specificities (Fig. 1B). Among
S1 family members, these loops display variable
lengths, with HGFA falling comfortably within the
ranges among close homologs. Additionally, some of
these loops (the 140-, 180-, and 220-loop) undergo con-
formational rearrangements during the zymogen to
enzyme transition, and, together with the N-terminal
peptide, are referred to as the ‘activation domain’ [8].
A zymogen form of ‘short HGFA’ has not been crystal-
lized, but we presume that HGFA undergoes analogous
changes during activation. A rare free Cys at posi-
tion 187 is not found in any close homolog, but seems
to have no special function [4]. Other key attributes of

the catalytically competent structure are the ‘oxyanion
hole’ formed by the amide nitrogens of Ser195 and
Gly193, and substrate-binding subsites (S1, S2, S3, and
S4), which interact with the corresponding P1–P4 resi-
dues of the substrate (Lys-Gln-Leu-Arg for pro-HGF)
(Fig. 1B). The principal determinant of substrate pref-
erence is the substrate-binding pocket, S1 (Fig. 1B). As
in trypsin, the Asp189 at the bottom of S1 confers a
strong preference for substrates with an Arg or Lys as
their P1 residue. In agreement with this, the two known
macromolecular substrates, pro-HGF and pro-MSP [2],
as well as synthetic substrates of HGFA, have a P1 Arg
residue [7,9].
Nonstandard active site conformations
The apo structure of HGFA (without inhibitor pres-
ent) reveals an active site in which key elements of the
substrate-binding site are changed in a way that is
incompatible with substrate binding and catalytic
activity. It shows a significant displacement of the
Ser214–Asp217 segment (part of the 220-loop) as com-
pared with the competent conformation with an ‘open’
active site. In apo-HGFA, the Ca atom of Trp215 is
shifted by 2.8 A
˚
and that of Gly216 by 5.5 A
˚
(Fig. 1C). As a consequence of this difference, the
entry of the substrate P1 residue into S1 is blocked,
and the active site is ‘closed’. Figure 1C shows that
this ‘closed’ conformation would cause a steric clash

with the P1 Arg, thus precluding a productive interac-
tion of a substrate with the catalytic machinery of
HGFA. The unconventional 220-loop arrangement is
supported by new hydrogen bonds and hydrophobic
interactions involving Trp215, also including some
interactions from a crystal packing contact.
This nonstandard 220-loop arrangement is not
limited to HGFA, but has precedents in the apo forms
of several other trypsin-like serine proteases, such as
Na
+
-free thrombin [10], a1-tryptase [11], tonin [12],
bacterial DegS [13], horse prostate kallikrein [14],
and prostasin [15,16]. In addition, on the basis of its
C. Eigenbrot et al. Molecular interactions of HGFA with inhibitor ⁄ substrate
FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS 2217
zymogen structure, Hink-Schauer et al. [17] postulated
that granzyme K may also have an S1 pocket
obstructed by a distorted 220-loop. All of these struc-
tures display significant displacement of residues 215–
220 and (usually) a few subsequent residues, although
there is not a strong consensus for a single ‘alterna-
tive ⁄ incompetent’ position for residues 215–220, per-
haps partly because of the influence of crystal packing
in some of these structures. Other features seen among
incompetent active sites are not observed for HGFA,
e.g. the loss of the oxyanion hole seen in Na
+
-free
thrombin. Among these proteases, the closest homolog

to HGFA is prostasin, also known as PRSS8 or chan-
nel-activating protease-1. In apo-prostasin, the 220-
loop is rearranged in a very similar manner as in apo-
HGFA, and obstructs substrate access to S1 (Fig. 1C)
[15,16]. As compared with the competent form of pro-
stasin, the 220-loop Asp217 is shifted by 5.4 A
˚
towards
S1 (Fig. 1C), and this new conformation is stabilized
by a network of hydrogen bonds to a network of
water molecules close to the catalytic Ser [15].
Moreover, the HGFA 99-loop can also adopt an
unconventional conformation, which is incompatible
with optimal enzyme activity. The 99-loop is important
for substrate–inhibitor interactions, as it contributes to
the formation of S2 and S4. The unconventional con-
formation affects the proper interaction of the sub-
strate with S2, owing to a rearrangement of the
99-loop residues Pro99a and Ser99, both of which
shape the relatively hydrophobic S2 (Fig. 1D) [9].
Due to the 99-loop movement, S2 is now smaller,
and interaction with the Leu P2 residue of substrates
is significantly impaired. This particular 99-loop con-
formation was observed in the structure of the inhibi-
tory Fab40 bound to HGFA, reflecting the mechanism
by which catalytic activity is inhibited by Fab40 [9].
Fab40 binds to a region outside of the substrate-bind-
ing cleft located at the ‘back side’ of the 99-loop, and
is a competitive allosteric inhibitor of HGFA [9]. It is
possible that the 99-loop ‘switch’ reflects a natural reg-

ulatory mechanism, as part of the Fab40-binding site
corresponds to thrombin exosite II, which is a known
effector-binding site regulating thrombin enzymatic
activity [18]. The apparent 99-loop conformational
flexibility is not restricted to HGFA, but has also been
observed or implied to occur in other S1 family mem-
bers, such as the close structural homolog prostasin,
where the 99-loop can adopt three different conforma-
tional states [16]. Members of the kallikrein family,
such as horse prostate kallikrein, have relatively long
99-loops, and in some structures with no substrate
mimic bound, the loop extends over the catalytic triad
and would restrict access by substrate [14]. In coagula-
tion factor IXa, the side chain of Tyr99 occludes S2 in
the absence of a substrate mimic [19].
In addition to the 99-loop, other substrate specific-
ity-determining loops, such as the 38-loop, 60-loop,
and 170-loop, can adopt different conformations in
various S1 family members [13,20,21], suggesting
remarkable plasticity of the serine protease active site
region. Thus, it is reasonable to assume that the
unconventional 99-loop and 220-loop conformations
of HGFA are part of an ensemble of conformational
states, and that the substrate is simply sampling the
conventional conformation, in effect shifting the equi-
librium towards the competent state.
It is common for structural studies of proteases to
include an inhibitor to limit autolysis and stabilize the
protein during crystallization. The characterization of
incompetent active sites among the relatively small

number of uninhibited S1 family X-ray structures sug-
gests that such conformational plasticity is widespread
and forms part of the biological regulation of enzyme
activity. Examples of regulation also include Na
+
effects on thrombin [10], the PDZ domain of DegS
[13], and interactions between the 220-loop and
99-loop of horse prostate kallikrein [14], and Ca
2+
effects on prostasin [16]. For systems without factors
beyond the protease domain playing a role, the free
energy requirement for moving between incompetent
and competent conformations is probably quite low,
well within the energy provided by substrate interac-
tions, and consistent with the notion of ‘induced fit’.
This is probably the case for HGFA, which, despite
adopting a catalytically incompetent conformation, is
enzymatically fully active. This suggests that HGFA
can easily undergo transition between the two active
site conformations. A contrary example is found for
the catalytically inactive a1-tryptase, which adopts an
incompetent 220-loop stabilized by a unique sequence
[11].
Inhibition of HGFA by HAI-1
The activity of HGFA is inhibited by naturally occur-
ring protein inhibitors belonging to different classes,
such as the Kunitz domain inhibitors HAI-1 and HAI-
2, and the serpin protein C inhibitor (SerpinA5) [22]
(refer to the review by Suzuki [23]). The first identified
inhibitor was HAI-1 [24], which is composed of an

N-terminal MANSC domain [25], a structurally unas-
signed region, KD1, low-density lipoprotein (LDL)
receptor-like domain, Kunitz domain-2 (KD2), a trans-
membrane domain, and a cytoplasmic domain
(Fig. 1A). A human splice variant containing an extra
16 amino acids inserted between KD1 and the LDL
Molecular interactions of HGFA with inhibitor ⁄ substrate C. Eigenbrot et al.
2218 FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS
receptor-like domain was identified, and was found to
have identical tissue expression and inhibitory specific-
ity to those of HAI-1 [26] (Fig. 2A). HAI-1 deficiency
is embryo lethal, owing to defective placental tissue
architecture caused by dysregulated enzymatic activity
[27–29]. KD1 inhibits the enzymatic activity of HGFA,
matriptase, hepsin, and prostasin, whereas the C-termi-
nal KD2 does not [26,30–33]. Domain characterization
studies suggest a complex interplay between various
HAI-1 domains in regulating the activity of KD1
towards the examined proteases, HGFA and matrip-
tase [30,32]. In the case of matriptase, the reported K
i
values range from 647 pm for the entire HAI-1 extra-
cellular domain to 1.6 pm for a smaller KD1-contain-
ing HAI-1 fragment [32]. The different binding
affinities of the full-length and truncated HAI-1 ver-
sions may be of physiological relevance, as several sol-
uble HAI-1 forms were found in the cell culture
medium [24] and in association with matriptase in
human milk [34].
HAI-1 inhibits HGFA by forming a tight associa-

tion as a pseudosubstrate between its KD1 and the
enzyme active site [4]. KD1 makes contacts (hydrogen
bonding and hydrophobic) to all substrate–inhibitor
specificity-determining loops of HGFA (Fig. 2A,B).
The conformations of these loops are essentially the
same as found in all examples of catalytically compe-
tent HGFA structures. There is a close correspondence
between the KD1 interactions with the substrate-bind-
ing cleft and those seen for the substrate mimic
KQLR-cmk. KD1 places the side chains of residues
Arg260, Cys259-Cys283, and Arg258 in the S1, S2 and
S4 subsites in a way almost identical to the way that
the substrate mimic KQLR places its P1, P2 and P4
side chains (Fig. 3). In addition, KD1 makes two main
chain to main chain hydrogen bonds with Ser214 and
Gly216 that are also formed by the substrate mimic
KQLR-cmk (Fig. 3A,B).
KD1 is one of the ‘standard mechanism’ or
‘Laskowski mechanism’ inhibitors, which tightly bind
to the enzyme in a substrate-like manner but undergo
cleavage at an extremely low rate. Indeed, the structure
shows KD1 presenting its intact P1–P1¢ (Arg-Gly)
peptide bond for nucleophilic attack by Ser195, the P1
backbone carbonyl being stabilized by the main chain
amide nitrogen atoms of Gly193 and Ser195 (Fig. 3B).
On the basis of biochemical and structural studies on a
related Kunitz domain–enzyme pair, the bovine
AB
Fig. 2. Interaction of HGFA protease domain with HAI-1-derived KD1 (PDB 1YC0). (A) KD1 (magenta) interacts with HGFA (beige) in a sub-
strate-like manner by occupying subsites S4, S2 and S1 (in orange) with Arg258, Cys259–Cys283, and Arg260 (P1 residue), respectively (side

chains in blue). The binding region is delineated by the dotted line, and corresponds to the green surface in (B). (B) Open book representa-
tion of the HGFA–KD1 interaction. Residues on HGFA (green) and KD1 (blue) with an atom within 4.0 A
˚
of the other protein (= binding
region) are indicated. KD1 makes contact with all substrate ⁄ inhibitor specificity-determining loops on HGFA (compare with Fig. 1B), and uses
the protruding P1 Arg260 for insertion into the deep S1 pocket. The catalytic His57 and Ser195 are also within 4.0 A
˚
of KD1 and are in
yellow.
C. Eigenbrot et al. Molecular interactions of HGFA with inhibitor ⁄ substrate
FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS 2219
pancreatic trypsin inhibitor–trypsin complex, it was
proposed that cleavage at the inhibitor P1–P1¢ peptide
bond can readily occur. However, owing to the tight
association of the cleavage product, the peptide bond
is more rapidly resynthesized, so that the intact form
of the Kunitz domain inhibitor predominates in the
crystal structure [35]. According to this model, the
intact KD1 peptide bond in the KD1–HGFA structure
thus reflects the capture of the dominant form, owing
to the more favorable rate of peptide bond resynthesis
during crystallization.
Enzymatic kinetic experiments showed that the
enzyme specificity of HAI-1 is completely determined
by KD1 alone, and does not require additional interac-
tions [4], although other HAI-1 domains may nega-
tively regulate the affinity of binding between KD1
and HGFA [30,32]. Therefore, the specificity must
arise from structural features of each inhibited enzyme
(e.g. HGFA, matriptase, hepsin, prostasin, and tryp-

sin) [4,8,15,36,37] around the inhibitor binding site.
The use of the KD1–HGFA complex to rationalize the
structural basis of enzyme specificity has obvious limi-
tations, as other KD1–enzyme structures are not avail-
able. Also, structural adjustments made by the enzyme
can be significant and difficult to predict, as in the case
of the related aprotinin–prostasin complex, in which
the 99-loop moves away from the substrate-binding
cleft to accommodate the Kunitz-type inhibitor aproti-
nin [15]. In some cases, however, good structural argu-
ments can be made, such as the complete lack of
inhibition by KD1 of the closely related uPA. A likely
reason is the conformation of the uPA 99-loop.
Although it is only one amino acid longer than the 99-
loop in HGFA, its conformation is quite different and,
in a hypothetical complex, it would extend well into
the location where Arg258 is found in HGFA S4,
causing a steric conflict with Leu97b of uPA
(PDB 1LMW). For a more detailed structure-based
analysis, see a previously published study by Shia et al.
[4].
Substrate interaction and specificity
The structures of the substrate mimic KQLR-cmk
bound to HGFA and the KD1–HGFA complex pro-
vide insights into salient features determining substrate
interactions and specificity. The KQLR peptide consti-
tutes the P4–P1 sequence of the natural substrate pro-
HGF, and thus should serve as a good approximation
of natural substrate interactions with the HGFA active
site. The KQLR-cmk peptide, covalently linked to the

catalytic Ser195 and His57, inserts into the active site
groove in a manner that is typical for substrate inter-
actions with trypsin-like serine proteases. It adopts a
twisted antiparallel conformation, forming the inter-
main chain hydrogen bonds between P1 Arg and
Ser214 and between P3 Gln and Gly216 (Fig. 3A). S1
is filled with the P1 Arg, which engages in standard
salt bridge interactions with HGFA Asp189, located at
the bottom of S1. The P2 Leu tightly packs into the
ABC
Fig. 3. Substrate interaction with HGFA. (A) The crystal structure of HGFA (surface representation, beige, PDB 2WUC) in complex with
Ac-KQLR-cmk (stick representation, green). The KQLR inhibitor is covalently bonded to Ser195 and His57, and it is stabilized by two inter-
main chain (P1 ArgÆSer214 and P3 GlnÆGly216) hydrogen bonds (red dotted lines). Additional hydrogen bonds with side chains include P1
ArgÆGly193, P2 LeuÆGln192, and P4 LysÆSer99. The hydrophobic S2 pocket is formed by His57, Ser99, Pro99a and Trp215 (orange). (B) The
structure of HGFA (surface representation, beige, PDB 1YC0) in complex with KD1 (stick representation, magenta). The Arg260 is bound in
the deep S1 pocket, and forms a salt bridge with Asp189 in a similar manner to the P1 Arg of KQLR. The carbonyl oxygen of Arg260 is
hydrogen bonded to the amide nitrogens of the oxyanion hole (Gly193 and Ser195). The hydrophobic S2 pocket (formed by His57, Ser99,
Pro99a, and Trp215) (blue) is occupied by thiols of disulfide-bonded Cys259–Cys283. Apart from forming a hydrogen bond with Ser99,
Arg258 of KD1 interacts with Trp215 via a p-stacking interaction. (C) Superposition of KQLR with the KD1 residues Arg258-Cys259 ⁄ Cys283-
Arg260 indicates an overlap of main chains P1–P3 and excellent correspondence of the side chains occupying subsites S1, S2, and S4.
Molecular interactions of HGFA with inhibitor ⁄ substrate C. Eigenbrot et al.
2220 FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS
small hydrophobic S2 pocket formed by Pro99a,
Ser99, Trp215, and His57, suggesting a strong prefer-
ence for a Leu at P2. The P3 Gln points outward
towards the solvent-exposed region of the active site,
suggesting poor specificity at this position, as in most
S1 family proteases. Some degree of specificity for S4
is suggested by the hydrogen bond formation between
P4 Lys and Ser99. Most intriguingly, the occupancy of

S1, S2 and S4 is reprised by the KD1 inhibitor, using
its Arg260 side chain, the thiol groups from the disul-
fide bonded Cys259–Cy283, and the Arg258 side chain,
respectively (Fig. 3B,C). This remarkable correspon-
dence may indicate that HGFA has preference for a
P1 Arg and a basic P4 (Lys ⁄ Arg) residue. Molecular
modeling studies indicate that a P4 Arg of the hypo-
thetical RQLR peptide would be an excellent fit, as it
may compensate for the negative electrostatic potential
created in S4 by Asp217 and Ser99 (data not shown).
Our structural arguments about the S4 specificity need
to be tempered by the facts that, for most S1 prote-
ases, the degree of specificity generally diminishes
beyond S2, and that our analysis is based on only two
crystal structures. In addition, the presence of a P2
Leu in both macromolecular and synthetic substrates
of HGFA is consistent with the structural features of
S2, suggesting a strong preference for Leu as a P2 resi-
due. The P1¢ residue for HAI-1 is a Gly, whereas it is
a Val for both known macromolecular substrates. This
may indicate a preference for small hydrophobic
residues at this position.
In conclusion, the intriguing structural features of
HGFA interactions with a substrate mimic and the
pseudosubstrate KD1 suggest that HGFA has unique
substrate preferences. This may be helpful in identify-
ing additional macromolecular substrates. Addition-
ally, the noncanonical conformations that have been
seen among HGFA protease structures may be useful
in discovering highly specific peptidic and nonpeptidic

inhibitors of HGFA.
References
1 Miyazawa K, Shimomura T, Kitamura A, Kondo J,
Morimoto Y & Kitamura N (1993) Molecular cloning
and sequence analysis of the cDNA for a human serine
protease reponsible for activation of hepatocyte growth
factor. Structural similarity of the protease precursor to
blood coagulation factor XII. J Biol Chem 268, 10024–
10028.
2 Kawaguchi M, Orikawa H, Baba T, Fukushima T &
Kataoka H (2009) Hepatocyte growth factor activator
is a serum activator of single-chain precursor macro-
phage-stimulating protein. FEBS J 276, 3481–3490.
3 Kataoka H & Kawaguchi M (2010) Hepatocyte growth
factor activator (HGFA): pathophysiological functions
in vivo. FEBS J 277, 2230–2237.
4 Shia S, Stamos J, Kirchhofer D, Fan B, Wu J, Corpuz
RT, Santell L, Lazarus RA & Eigenbrot C (2005) Con-
formational lability in serine protease active sites: struc-
tures of hepatocyte growth factor activator (HGFA)
alone and with the inhibitory domain from HGFA
inhibitor-1B. J Mol Biol 346, 1335–1349.
5 Schechter I & Berger A (1968) On the active site of pro-
teases. 3. Mapping the active site of papain; specific
peptide inhibitors of papain. Biochem Biophys Res
Commun 32, 898–902.
6 Perona JJ & Craik CS (1995) Structural basis of sub-
strate specificity in the serine proteases. Protein Sci 4,
337–360.
7 Wu Y, Eigenbrot C, Liang WC, Stawicki S, Shia S,

Fan B, Ganesan R, Lipari MT & Kirchhofer D (2007)
Structural insight into distinct mechanisms of protease
inhibition by antibodies. Proc Natl Acad Sci USA 104,
19784–19789.
8 Huber R & Bode W (1978) Structural basis of the acti-
vation and action of trypsin. Acc Chem Res 11, 114–
122.
9 Ganesan R, Eigenbrot C, Wu Y, Liang W-C, Shia S,
Lipari MT & Kirchhofer D (2009) Unraveling the allo-
steric mechanism of serine protease inhibition by an
antibody. Structure 17, 1614–1624.
10 Johnson DJ, Adams TE, Li W & Huntington JA (2005)
Crystal structure of wild-type human thrombin in the
Na
+
-free state. Biochem J 392, 21–28.
11 Marquardt U, Zettl F, Huber R, Bode W &
Sommerhoff C (2002) The crystal structure of human
alpha1-tryptase reveals a blocked substrate-binding
region. J Mol Biol 321, 491–502.
12 Fujinaga M & James MN (1987) Rat submaxillary
gland serine protease, tonin. Structure solution and
refinement at 1.8 A resolution. J Mol Biol 195,
373–396.
13 Wilken C, Kitzing K, Kurzbauer R, Ehrmann M &
Clausen T (2004) Crystal structure of the DegS stress
sensor: how a PDZ domain recognizes misfolded pro-
tein and activates a protease. Cell 117, 483–494.
14 Carvalho AL, Sanz L, Barettino D, Romero A, Calvete
JJ & Romao MJ (2002) Crystal structure of a prostate

kallikrein isolated from stallion seminal plasma: a
homologue of human PSA. J Mol Biol 322, 325–337.
15 Rickert KW, Kelley P, Byrne NJ, Diehl RE, Hall DL,
Montalvo AM, Reid JC, Shipman JM, Thomas BW,
Munshi SK et al. (2008) Structure of human prostasin,
a target for the regulation of hypertension. J Biol Chem
283, 34864–34872.
16 Spraggon G, Hornsby M, Shipway A, Tully DC, Bursu-
laya B, Danahay H, Harris JL & Lesley SA (2009)
Active site conformational changes of prostasin provide
C. Eigenbrot et al. Molecular interactions of HGFA with inhibitor ⁄ substrate
FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS 2221
a new mechanism of protease regulation by divalent
cations. Protein Sci 18, 1081–1094.
17 Hink-Schauer C, Estebanez-Perpina E, Wilharm E, Fu-
entes-Prior P, Klinkert W, Bode W & Jenne DE (2002)
The 2.2-A crystal structure of human pro-granzyme K
reveals a rigid zymogen with unusual features. J Biol
Chem 277, 50923–50933.
18 Bode W (2005) The structure of thrombin, a chame-
leon-like proteinase. J Thromb Haemost 3, 2379–2388.
19 Hopfner KP, Lang A, Karcher A, Sichler K, Kopetzki E,
Brandstetter H, Huber R, Bode W & Engh RA (1999)
Coagulation factor IXa: the relaxed conformation of
Tyr99 blocks substrate binding. Structure 7, 989–996.
20 van de Locht A, Bode W, Huber R, Le Bonniec BF,
Stone SR, Esmon CT & Stubbs MT (1997) The thrombin
E192Q–BPTI complex reveals gross structural rearrange-
ments: implications for the interaction with antithrombin
and thrombomodulin. EMBO J 16, 2977–2984.

21 Li W, Adams TE, Nangalia J, Esmon CT & Hunting-
ton JA (2008) Molecular basis of thrombin recognition
by protein C inhibitor revealed by the 1.6-A structure
of the heparin-bridged complex. Proc Natl Acad Sci
USA 105, 4661–4666.
22 Hayashi T, Nishioka J, Nakagawa N, Kamada H, Gab-
azza EC, Kobayashi T, Hattori A & Suzuki K (2007)
Protein C inhibitor directly and potently inhibits acti-
vated hepatocyte growth factor activator. J Thromb
Haemost 5, 1477–1485.
23 Suzuki K (2010) Hepatocyte growth factor activator
(HGFA): its regulation by protein C inhibitor. FEBS J
277, 2223–2229.
24 Shimomura T, Denda K, Kitamura A, Kawaguchi T,
Kito M, Kondo J, Kagaya S, Qin L, Takata H, Miyaz-
awa K et al. (1997) Hepatocyte growth factor activator
inhibitor, a novel Kunitz-type serine protease inhibitor.
J Biol Chem 272, 6370–6376.
25 Guo J, Chen S, Huang C, Chen L, Studholme DJ,
Zhao S & Yu L (2004) MANSC: a seven-cysteine-con-
taining domain present in animal membrane and extra-
cellular proteins. Trends Biochem Sci 29, 172–174.
26 Kirchhofer D, Peek M, Li W, Stamos J, Eigenbrot C,
Kadkhodayan S, Elliott JM, Corpuz RT, Lazarus RA
& Moran P (2003) Tissue expression, protease specific-
ity, and Kunitz domain functions of hepatocyte growth
factor activator inhibitor-1B (HAI-1B), a new splice
variant of HAI-1. J Biol Chem 278, 36341–36349.
27 Szabo R, Molinolo A, List K & Bugge TH (2007) Ma-
triptase inhibition by hepatocyte growth factor activator

inhibitor-1 is essential for placental development. Onco-
gene 26, 1546–1556.
28 Fan B, Brennan J, Grant D, Peale F, Rangell L &
Kirchhofer D (2007) Hepatocyte growth factor activator
inhibitor-1 (HAI-1) is essential for the integrity of
basement membranes in the developing placental
labyrinth. Dev Biol 303, 222–230.
29 Tanaka H, Nagaike K, Takeda N, Itoh H, Kohama K,
Fukushima T, Miyata S, Uchiyama S, Uchinokura S,
Shimomura T et al. (2005) Hepatocyte growth factor
activator inhibitor type 1 (HAI-1) is required for
branching morphogenesis in the chorioallantoic pla-
centa. Mol Cell Biol 25 , 5687–5698.
30 Denda K, Shimomura T, Kawaguchi T, Miyazawa K
& Kitamura N (2002) Functional characterization
of Kunitz domains in hepatocyte growth factor
activator inhibitor type 1. J Biol Chem 277,
14053–14059.
31 Kirchhofer D, Peek M, Lipari MT, Billeci K, Fan B &
Moran P (2005) Hepsin activates pro-hepatocyte growth
factor and is inhibited by hepatocyte growth factor acti-
vator inhibitor-1B (HAI-1B) and HAI-2. FEBS Lett
579, 1945–1950.
32 Kojima K, Tsuzuki S, Fushiki T & Inouye K (2008)
Roles of functional and structural domains of hepato-
cyte growth factor activator inhibitor type 1 in the inhi-
bition of matriptase. J Biol Chem 283 , 2478–2487.
33 Fan B, Wu TD, Li W & Kirchhofer D (2005) Identifica-
tion of hepatocyte growth factor activator inhibitor-1B
as a potential physiological inhibitor of prostasin. J Biol

Chem 280, 34513–34520.
34 Lin CY, Anders J, Johnson M & Dickson RB (1999)
Purification and characterization of a complex contain-
ing matriptase and a Kunitz-type serine protease inhibi-
tor from human milk. J Biol Chem 274, 18237–18242.
35 Zakharova E, Horvath MP & Goldenberg DP (2009)
Structure of a serine protease poised to resynthesize a
peptide bond. Proc Natl Acad Sci USA 106, 11034–
11039.
36 Friedrich R, Fuentes-Prior P, Ong E, Coombs G, Hun-
ter M, Oehler R, Pierson D, Gonzalez R, Huber R,
Bode W et al. (2002) Catalytic domain structures of
MT-SP1 ⁄ matriptase, a matrix-degrading transmembrane
serine proteinase. J Biol Chem 277, 2160–2168.
37 Somoza JR, Ho JD, Luong C, Ghate M, Sprengeler
PA, Mortara K, Shrader WD, Sperandio D, Chan H,
McGrath ME et al. (2003) The structure of the extracel-
lular region of human hepsin reveals a serine protease
domain and a novel scavenger receptor cysteine-rich
(SRCR) domain. Structure 11, 1123–1131.
38 DeLano WL. The PyMOL Molecular Graphics Systems,
2002.
Molecular interactions of HGFA with inhibitor ⁄ substrate C. Eigenbrot et al.
2222 FEBS Journal 277 (2010) 2215–2222 ª 2010 The Authors Journal compilation ª 2010 FEBS