Báo cáo khoa học: Binding areas of urokinase-type plasminogen activator– plasminogen activator inhibitor-1 complex for endocytosis receptors of the low-density lipoprotein receptor family, determined by site-directed mutagenesis doc

Binding areas of urokinase-type plasminogen activator–
plasminogen activator inhibitor-1 complex for endocytosis
receptors of the low-density lipoprotein receptor family,
determined by site-directed mutagenesis
Sune Skeldal
1
, Jakob V. Larsen
1
, Katrine E. Pedersen
1
, Helle H. Petersen
1
, Rikke Egelund
1
,
Anni Christensen
1
, Jan K. Jensen
1,2
, Jørgen Gliemann
3
and Peter A. Andreasen
1,2
1 Department of Molecular Biology, University of Aarhus, Denmark
2 Interdisciplinary Nanoscience Center (iNANO), University of Aarhus, Denmark
3 Department of Medical Biochemistry, University of Aarhus, Denmark
The low-density lipoprotein receptor (LDLR) family of
endocytosis receptors has been implicated in binding
and endocytosis of a large number of structurally un-
related proteins, including apolipoproteins, protease–
inhibitor complexes, extracellular matrix proteins, and

hormone carriers. In mammals, this receptor family
includes LDLR itself, low-density lipoprotein receptor-
related protein-1A (LRP-1A), LRP-1B, megalin
or LRP-2, very-low-density lipoprotein receptor
(VLDLR), and apolipoprotein E receptor-2. These
Keywords
low-density lipoprotein receptor-related
protein; plasminogen activator inhibitor 1;
sorting protein-related receptor; urokinase
plasminogen activator; very-low-density
lipoprotein receptor
Correspondence
P. A. Andreasen, Department of Molecular
Biology, University of Aarhus, Gustav
Wied’s Vej 10C, 8000 Aarhus C, Denmark
Fax: +45 86 12 31 78
Tel: +45 89 42 50 80
E-mail:
(Received 17 July 2006, revised 20 Septem-
ber 2006, accepted 22 September 2006)
doi:10.1111/j.1742-4658.2006.05511.x
Some endocytosis receptors related to the low-density lipoprotein receptor,
including low-density lipoprotein receptor-related protein-1A, very-low-
density lipoprotein receptor, and sorting protein-related receptor, bind pro-
tease-inhibitor complexes, including urokinase-type plasminogen activator
(uPA), plasminogen activator inhibitor-1 (PAI-1), and the uPA–PAI-1 com-
plex. The unique capacity of these receptors for high-affinity binding of
many structurally unrelated ligands renders mapping of receptor-binding
surfaces of serpin and serine protease ligands a special challenge. We have
mapped the receptor-binding area of the uPA–PAI-1 complex by site-direc-

ted mutagenesis. Substitution of a cluster of basic residues near the 37-loop
and 60-loop of uPA reduced the receptor-binding affinity of the uPA–PAI-1
complex approximately twofold. Deletion of the N-terminal growth factor
domain of uPA reduced the affinity 2–4-fold, depending on the receptor,
and deletion of both the growth factor domain and the kringle reduced the
affinity sevenfold. The binding affinity of the uPA–PAI-1 complex to the
receptors was greatly reduced by substitution of basic and hydrophobic resi-
dues in a-helix D and a-helix E of PAI-1. The localization of the implicated
residues in the 3D structures of uPA and PAI-1 shows that they form a
continuous receptor-binding area spanning the serpin as well as the
A-chain and the serine protease domain of uPA. Our results suggest that
the 10–100-fold higher affinity of the uPA–PAI-1 complex compared with
the free components depends on the bonus effect of bringing the binding
areas on uPA and PAI-1 together on the same binding entity.
Abbreviations
a
1
-PI, a
1
-antiproteinase inhibitor; CTR, complement type repeat; HEK293T, human embryonic kidney cell line 293T; LDLR, low-density
lipoprotein receptor; LRP, low-density lipoprotein receptor-related protein; PAI-1, plasminogen activator inhibitor 1; RAP, receptor-associated
protein; RCL, reactive centre loop; sorLA, sorting protein-related receptor; SPD, serine protease domain; tPA, tissue-type plasminogen
activator; uPA, urokinase-type plasminogen activator; uPAR, uPA receptor; VLDLR, very-low-density lipoprotein receptor.
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5143
receptors are constructed with the same types of
domains, but with variable numbers of each type of
domain. The domains include complement-type repeats
(CTRs), YWTD-repeat-containing b-propellers, and
epidermal growth factor precursor domains. The recep-
tors also all have a transmembrane a-helix and a cyto-

plasmic C-terminal domain mediating endocytosis via
clathrin-coated pits. Generally, the CTRs are believed
to mediate ligand binding. The related receptor sorting
protein-related receptor (sorLA), which in addition to
other types of domains also contains CTRs, have a lig-
and repertoire overlapping that of the LDLR family.
One ligand common to all these receptors is the
40-kDa receptor-associated protein (RAP) [1].
The crystal structures of the third domain of RAP
in complex with a CTR pair from LDLR and of
human rhinovirus serotype 2 in complex with CTRs
from VLDLR have recently been solved [2,3]. Com-
mon to all of these structures is the fact that binding
to the CTRs is mediated through basic and hydropho-
bic residues in the ligand. Also, some inferences can be
made from an X-ray structure analysis of LDLR crys-
tals obtained at pH 5, in which the b-propeller bends
back and makes contact with the CTR cluster in a
way believed to mimic ligand binding [4].
The fact that these receptors exhibit high-affinity
binding of so many structurally unrelated ligands
makes their ligand-binding potential a unique case of
molecular recognition. Particularly interesting ligands
are the serine protease urokinase-type plasminogen
activator (uPA), its primary serpin inhibitor, plasmino-
gen activator inhibitor-1 (PAI-1), and the correspond-
ing protease-serpin complex, which have been shown
to bind to LRP-1A [5], megalin [6,7], VLDLR [8,9],
LRP-1B [10], and sorLA [11]. These receptors mediate
endocytosis of uPA–PAI-1 complex accumulated on

the cell surface by binding to the urokinase-type
plasminogen activator receptor (uPAR), whereas endo-
cytosis directly from the fluid phase is negligible [5].
Although both free uPA and free PAI-1 exhibit a dis-
tinct affinity for these receptors, the uPA–PAI-1 com-
plex binds to LRP-1A with an affinity much higher
than that of the free components. Thus, the K
d
value
for binding of the uPA–PAI-1 complex to these recep-
tors is reported to be % 1nm [6,8,11,12], whereas that
for binding of free PAI-1 is reported to be % 30 nm
[11–13], and K
d
values of 7–200 nm have been reported
for binding of free u-PA [11,12,14].
The molecular recognition between the receptors
and the uPA–PAI-1 complex should therefore be seen
in relation to the mechanism of protease–serpin com-
plex formation. X-ray crystal structure analyses have
shown that serpins are globular proteins consisting of
three b-sheets and nine a-helices [15] (Fig. 1). Three-
dimensional structures of a covalently coupled a
1
-anti-
proteinase inhibitor (a
1
-PI)–trypsin complex [16], a
covalently coupled a
1

-PI–elastase complex [17], and
several reversible complexes between serpins and pro-
teases with the active-site Ser replaced by Ala [18–20]
have been reported. The structures support biochemi-
cal and biophysical evidence that complex formation is
initiated by formation of a reversible docking complex,
in which the P
1
–P
1
¢ bond in the surface-exposed react-
ive centre loop (RCL) interacts with the active site of
the protease. Next, the P
1
–P
1
¢ bond is cleaved, the P
1
residue coupled to the active-site Ser of the protease
by an ester bond, the N-terminal part of the RCL
inserted as strand 4 in b-sheet A, and the protease
translocated to the opposite pole of the serpin [15].
Serpins are thus attacked by the proteases as sub-
strates, but the normal catalytic cycle stops at the acyl-
enzyme intermediate stage. From the available 3D
structures of stable protease–serpin complexes [16,17],
it was inferred that the catalytic mechanism is halted
because of distortion of the active site of the protease.
The energy needed for the distortion stems from stabil-
ization of the serpin in the ‘relaxed’ conformation by

insertion of the RCL into b-sheet A, as opposed to the
‘stressed’, relatively unstable active conformation.
RCL insertion can also occur after abortive complex
formation as the result of complete cleavage of the
P
1
–P
1
¢ bond or by insertion of the uncleaved RCL in
latent PAI-1 [15].
Unfortunately, there is no X-ray crystal structure
analysis of any receptor–protease–inhibitor complexes.
However, on the basis of the established inhibitory
mechanism for serpins, there are several possible expla-
nations of the increased receptor-binding affinity of
the uPA–PAI-1 complex compared with the affinity of
the individual components, including increased affinity
associated with a conformational change in PAI-1,
increased affinity associated with a conformational
change in uPA, and ⁄ or an avidity effect of two or
more binding sites being brought together on the same
ligand. We have now addressed this problem using
site-directed mutagenesis, introducing mutations into
both uPA and PAI-1, and studying binding to
VLDLR, LRP-1A, and sorLA. We initially chose basic
residues for mutation on the basis of evidence for
receptor binding involving interactions between basic
residues in the ligands and acidic residues in the recep-
tor [21]. Binding of the uPA–PAI-1 complex to LRP-
1A and VLDLR was previously shown to require

basic residues in a-helix D and a-helix E in the flexible
joint-region of the inhibitor [13,22,23]. We
therefore mutated residues adjacent to this region in
uPA–PAI-1 complex receptor binding S. Skeldal et al.
5144 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
the uPA–PAI-1 complex, and we provide evidence that
the complex has an extended receptor-binding area
spanning uPA as well as PAI-1.
Results
Screening of the effects of Ala substitutions in
PAI-1 on uPA–PAI-1 complex–receptor binding in
microtiter wells
For the initial screening, our strategy was to substitute
basic and hydrophobic residues in PAI-1 with Ala,
prepare complexes between the variants and
125
I-uPA,
measure the binding of 20 pm radioactive complexes to
VLDLR, LRP-1A, or sorLA immobilized in microtiter
wells, and express the binding of the Ala-substituted
complexes relative to the binding of the wild-type com-
plex. This relatively simple strategy was based on
several facilitating arrangements.
We used PAI-1 expressed in a human cell line,
because of solubility problems with nonglycosylated,
bacterially expressed PAI-1. The source of PAI-1 for the
preparation of the complexes was not purified PAI-1,
but serum-free conditioned medium from human
embryonic kidney cell line 293T (HEK293T) transfected
with the corresponding cDNA. A well-established pro-

cedure [24] was used to re-activate the latent PAI-1 in
the medium by the use of denaturation with SDS and
refolding by removing the SDS by addition of an excess
of BSA.
125
I-uPA–PAI-1 complexes were prepared by
adding
125
I-uPA to the medium and purifying the com-
plexes by immuno-affinity chromatography [24].
As a control for the integrity of the
125
I-uPA–PAI-1
complexes during the binding assay, we measured the
binding of all Ala-substituted complexes to parallel
wells coated with 20 lgÆmL
)1
monoclonal antibody to
PAI-1 (mAb2). This antibody coat bound 64 ± 10%
of all complexes, and the binding of variant complexes
always varied less than 20% from the binding of the
wild-type complex.
Binding of the uPA–PAI-1 complex to the receptors
was assumed to follow the equation:
Fig. 1. The 3D structures of PAI-1 and uPA. The figure shows ribbon diagrams of the structure of the active conformation of PAI-1 [53] (pdb
file 1B3K) and the structure of the SPD of uPA [27] (pdb file 1LMW). The localization of several a-helices and b-strands are indicated. The
diagrams were constructed with the use of
SWISSPDBVIEWER.
S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5145

½RL¼½R
T
½L=ðK
d
þ½LÞ
in which [RL] is receptor-bound ligand, [R]
T
is total
receptor concentration, [L] is the concentration of free
ligand, and K
d
is the equilibrium dissociation constant.
With the 20 pm concentrations of
125
I-labeled ligands
used here and K
d
values % 1nm (see above),
[L] << K
d
, and the above equation is reduced to:
ð½RL=½LÞ ¼ ½R
T
=K
d
Hence, the fractional uPA–PAI-1 complex–receptor
binding is expected to be inversely proportional to the
K
d
value. Control experiments (not shown) confirmed

the linear relationship between [RL] and [L]. We there-
fore used the amount of receptor-bound complex
between
125
I-labelled uPA and mutated PAI-1, as com-
pared with the amount of receptor-bound complex
between
125
I-labeled uPA and wild-type PAI-1, as a
measure of the effect of the substitution on the K
d
value. With the accuracy and background binding of
this assay, we expected that binding of up to 10-fold
less than the control value would be different from
background.
For analyses of the binding of complexes of uPA
with PAI-1 mutants, we first mutated groups of
three residues to Ala. If a triple mutation resulted in
more than a twofold reduction in binding of the
uPA–PAI-1 complex to the receptor compared with
the wild-type complex, we investigated the effect of
substituting each residue separately. For single
mutants, a receptor-binding reduction of more than
1.5-fold was considered to define the residues
involved in receptor binding. Residues for Ala substi-
tutions were selected on the basis of their proximity
to the previously implicated residues in a-helix D
and a-helix E, namely Lys71, Arg78, Lys82, Lys90,
Arg120 and Lys124 [13,22,23].
With this strategy, out of a total of 41 PAI-1 resi-

dues tested, we identified 11 of importance for binding
of the uPA–PAI-1 complex to one, two, or all three
receptors (Fig. 2). However, the involvement of
Lys124 in receptor binding could not be confirmed.
Screening of effects of Ala substitutions in uPA
on uPA–PAI-1 complex–receptor binding in
microtiter wells
Besides the C-terminal, % 30-kDa serine protease
domain (SPD), uPA contains an N-terminal, % 25-
kDa A-chain, consisting of a growth factor domain
Fig. 2. Effect of Ala substitutions of PAI-1 residues on the binding of the uPA–PAI-1 complex to VLDLR, LRP-1A and sorLA in a solid-phase
assay. The binding of the variant complexes was expressed relative to the binding of wild-type complex in the same experiment. Mean ± SD
values are shown for at least three independent experiments. As compared with the binding of wild-type complexes, the binding to the
receptors of the complexes between uPA and the following PAI-1 mutants were reduced less than twofold when tested as triple mutants
or less than 1.5-fold when tested individually, and these mutated residues were therefore considered to be unimportant in the binding: H4A,
H5A; P6A; P7A; Y9A; Q58A; K67A; D69A; D70A; P75A; L77A; M85A; P87A; W88A; E92A; T96A; R103A; D104A; K106; L107A; Q109A;
P113A; H114; F119A; S121A; K124A; Q125A; W141A; H145A; K178A. °, Significantly different from binding observed with the correspond-
ing wild-type (P<0.01).
uPA–PAI-1 complex receptor binding S. Skeldal et al.
5146 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
and a kringle domain [25]. In agreement with previous
reports on LRP-1A [12] and sorLA [11], we now dem-
onstrate that a truncated version of uPA without the
A-chain (LMW-uPA) has a sevenfold reduced affinity
for VLDLR (Fig. 3). Moreover, we demonstrated that
a uPA variant with deletion of the growth factor
domain (DGF-uPA) had 2–4-fold reduced binding to
VLDLR, LRP-1A, and sorLA (see Fig. 3). Suspecting
the involvement of basic residues in the contact with
the receptors, we measured the effect on receptor bind-

ing of substituting a cluster of three arginines (R108,
R109, and R110) in the uPA kringle, but found no
change in receptor binding after these mutations
(Fig. 3). At the moment, possible endocytosis receptor-
binding residues in the kringle is not known. We did
not perform a mutational analysis of the growth factor
domain, as this, under physiological conditions at the
cell surface, is shielded from contact with the endo-
cytosis receptors by binding to uPAR [5,12].
We previously reported that a number of mono-
clonal antibodies against the SPD of uPA have epi-
topes localized in the 37-loop and 60-loop [26]
(Fig. 1). To preliminarily investigate whether a recep-
tor-binding site exists in the SPD of uPA, we
studied the effect on uPA–PAI-1 complex–receptor
binding of one such monoclonal antibody, mAb3689,
with an epitope encompassing Arg179(36),
His180(37), and Arg181(37a) in the 37-loop of uPA
(amino-acid residues in the SPD of uPA will be
referred to by a double numbering system, based on
numbering from the N-terminus of the native protein
with the chymotrypsin template numbering system in
parentheses; residues in the N-terminal A-chain of
uPA will be referred to by numbering from the
N-terminus of the native protein [27]). The presence
of the antibody reduced the binding significantly
(Fig. 3). We therefore studied the effect of Ala sub-
stitution of clusters of residues in the 37-loop and
60-loop on receptor binding and demonstrated that
both sets of mutations reduced receptor binding 2–4-

fold (Fig. 3). As a control, we found that mAb3689
did not reduce the binding of the variant with the
mutations in its epitope in the 37-loop (Fig. 3). Sub-
stitution of a number of other residues was without
effect on receptor binding (Fig. 3). We therefore con-
cluded that two or more residues in the 37-loop and
60-loop of the SPD of uPA are part of the ligand–
receptor interface.
Fig. 3. Effect of Ala substitutions of uPA
residues on the binding of the uPA–PAI-1
complex to VLDLR, LRP-1A, and sorLA in a
solid-phase assay. The binding of the variant
complexes was expressed relative to the
binding of the wild-type complex in the
same experiment. Mean ± SD values are
shown for at least three independent experi-
ments. Compared with binding of wild-type
complexes, binding to the receptors of the
complexes between PAI-1 and the following
uPA mutants were reduced less than
1.5-fold, and the mutated residues were
therefore considered unimportant for bind-
ing: K212(62)A; E213(62a)A; D194(63)A;
I216(65)A; Y218(67)A; N227(76)A;
Q229(78)A; E235(84)A; K264(110a)A;
E265(110b)A; R267(110d)A; H402(241)A;
K404(243)A. °, Significantly different from
binding observed with the corresponding
wild-type complex (P<0.01). ND, not deter-
mined.

S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5147
Screening of effects of Ala substitutions in both
uPA and PAI-1 on uPA–PAI-1 complex–receptor
binding in microtiter wells
To extend the results obtained by the microtiter well-
binding assays by Biacore binding analysis and assays
of receptor-mediated endocytosis, we used a complex
between a PAI-1 variant with low receptor affinity and
a uPA variant with low receptor affinity. We used a
PAI-1 variant with a quadruple mutation in a-helix D,
i.e. K71A-R78A-Y81A-K82A. This variant did not dif-
fer significantly from wild-type PAI-1 with respect to
specific inhibitory activity (wild-type 69 ± 4% and
mutant 59 ± 7% of the theoretical maximum),
second-order rate constant for the reaction with
uPA (wild-type 3.3 ± 0.7 · 10
6
m
)1
Æs
)1
and mutant
3.7 ± 0.5 · 10
6
m
)1
Æs
)1
), and relative vitronectin bind-

ing (mutant 1.03 ± 0.11 times that of wild-type). We
used the uPA variant with the triple mutation in the
37-loop, i.e. R178(35)A-R179(36)A-R181(37a)A. This
variant did not differ significantly from wild-type uPA
with respect to K
m
for hydrolysis of S-2444 (pyro-Glu-
Gly-Arg-p-nitroanilide), the K
m
values for wild-type
and variant being 87 ± 11 lm and 85 ± 0.01 lm,
respectively. The resulting complex showed the
expected more than 10-fold reduction in affinity for
VLDLR and LRP-1A in microtiter well-binding assays
(Fig. 4).
Surface plasmon resonance analysis of receptor
binding of uPA–PAI-1 complexes
We analysed receptor binding of the complexes
between the quadruple a-helix D PAI-1 mutant
(K71A-R78A-Y81A-K82A PAI-1) and the 37-loop
uPA mutant [R178(35)A-R179(36)A-R181(37a)A uPA]
by surface plasmon resonance. In this case, nonradio-
active wild-type and mutant uPA–PAI-1 complexes
were prepared from PAI-1 purified by immuno-affinity
chromatography and re-activated by denaturation with
guanidinium chloride and refolding by dialysis, and
uPA purified by immuno-affinity chromatography. The
complexes were purified from unreacted uPA and
PAI-1 by immuno-affinity chromatography. VLDLR,
LRP-1A, or sorLA was immobilized on Biacore chips,

and wild-type and mutant uPA–PAI-1 complexes injec-
ted on to the chips at concentrations of 1.5–100 nm.
The time course of the binding obtained with VLDLR
is shown in Fig. 5. Similar results were found with the
two other receptors. It is evident from the figure that
the binding was reduced with the mutant uPA–wild-
type PAI-1 complex, the wild-type uPA–mutant PAI-1
complex, and, in particular, with the mutant
uPA–mutant PAI-1 complex. Fitting the binding data
to a Langmuir 1 : 1 binding model by the use of the
biaevaluation 3.0 software (global fitting) resulted in
a K
d
value for binding of the wild-type uPA–wild-type
PAI-1 complex to either of the three receptors of
% 1nm, while K
d
for the mutant uPA–mutant PAI-1
complex was increased more than 10-fold. However,
the fit to the simple 1 : 1 binding model was poor. As
appears from Fig. 5, the rapid first phase of the associ-
ation was followed by a second, slower phase, and a
corresponding rapid dissociation phase, amounting to
20% of total binding, particularly with relatively high
ligand concentrations. We therefore carried out a more
Fig. 4. Effect of Ala substitutions of both uPA and PAI-1 residues
on binding of the uPA–PAI-1 complex to VLDLR, LRP-1A, and sor-
LA in a solid-phase assay. Binding of wild-type and variant com-
plexes to the receptors was estimated. Binding of the variant
complexes is expressed relative to that of the wild-type complex in

the same experiment. Mean ± SD values are shown for at least
three independent experiments. °, Significantly different from bind-
ing observed with the corresponding wild-type (P<0.01). ND, not
determined.
uPA–PAI-1 complex receptor binding S. Skeldal et al.
5148 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
reliable estimation of the K
d
values by determining the
association rate constants from the initial rate of
association and the dissociation rate constant from the
slow phase of dissociation (see Experimental proce-
dures) and calculated the K
d
values as the ratio
between the association and dissociation rate constants
(Table 1). The K
d
values obtained for the wild-type
uPA–wild-type PAI-1 complex, %1nm, were in good
agreement with those reported previously [6,8,11,12].
The K
d
values of the complexes between mutant uPA
and wild-type PAI-1 and wild-type uPA and mutant
PAI-1 were increased 1.5–5-fold, mostly as the result
of increased dissociation rate constants. A particularly
large change, corresponding to 10–30-fold increased K
d
values, was observed for the complex with mutations

in both uPA and PAI-1. This increase was the result
of a decreased association rate constant and an
increased dissociation rate constant. The fold reduction
in the K
d
values determined by surface plasmon reson-
ance did not differ significantly from those expected
from the microtiter well-binding assays, although the
average fold reductions of the K
d
value were smaller in
the Biacore experiments than expected from the
microtiter well-binding assays.
Effects of Ala substitutions of uPA and PAI-1 on
receptor-mediated endocytosis of the uPA–PAI-1
complex
We measured the receptor-dependent degradation of
the complexes between the quadruple a-helix D PAI-1
mutant (K71A-R78A-Y81A-K82A PAI-1) and the
37-loop uPA mutant [R178(35)A-R179(36)A-
R181(37a)A uPA] in cell lines expressing VLDLR and
LRP-1A. For VLDLR-mediated endocytosis, we used
U937 cells. These cells were previously shown to con-
tain VLDLR-II mRNA, i.e. a VLDLR variant without
exon 16 encoding the O-linked sugar domain [28]. Lig-
and blot analysis of membrane fragments from U937
cells revealed a RAP-binding membrane protein
co-migrating with VLDLR-II (Fig. 6). For LRP-1A-
mediated endocytosis, we used COS-1 cells, in which
the only RAP-binding receptor detectable by RAP lig-

and blotting analysis is LRP-1A [29]. In both cell lines,
Fig. 5. Surface plasmon resonance analysis of binding of wild-type
and variant uPA–PAI-1 complexes to VLDLR. Binding was meas-
ured using chips with % 50 fmol ⁄ mm
2
immobilized VLDLR. The
chips were superfused with the indicated complexes at concentra-
tions of 12, 6, 3 and 1.5 n
M, followed by buffer alone at 480 s.
Mutant uPA ¼ R178(35)A-R179(36)A-R181(37a)A uPA. Mutant
PAI-1 ¼ K71A-R78A-Y81A-K82A PAI-1.
S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5149
the receptor-mediated degradation of wild-type and
variant complexes was reduced largely in parallel with
receptor binding (Fig. 6).
Using another PAI-1 mutant, i.e. a triple mutant con-
taining three different substitutions each resulting in
reduced receptor binding, we also showed that reduced
complex–receptor binding was associated with increased
accumulation of the complexes on the cell surface,
compared with the wild-type complex. Another triple
mutant without reduced receptor binding did not result
in accumulation on the cell surface (Table 2).
It is not possible to study the endocytosis of variant
complexes with truncations of the A-chain of uPA.
These variants do not bind to uPAR and their endo-
cytosis is therefore negligible.
Discussion
In this work, we used site-directed mutagenesis to map

the VLDLR, LRP-1A, and sorLA binding surfaces of
the uPA–PAI-1 complex. The mapped interaction sur-
face spans both PAI-1 and uPA. In PAI-1, residues
His79, Tyr81, Met112, Phe116, Arg117, and Arg270
are implicated in the interaction surface. Also, the pre-
viously reported reduced binding of the double
mutants K82A-R120A and R78A-K124A [22] could be
tracked back to reveal the importance of Arg120, as
well as Lys82 and Arg78, in receptor binding, and con-
firm the previous reports of the importance of Lys71,
Lys82, and Lys90 [13] and Arg78 [23] in a-helix D. Of
the 41 PAI-1 residues tested because of their proximity
to a-helix D, 12 were Arg or Lys. Of the 11 residues
implicated in binding, seven were Arg or Lys. Thus,
basic residues constituted a much higher percentage of
Table 1. Effect of Ala substitutions of uPA and PAI-1 residues on the binding of uPA–PAI-1 complex to VLDLR, LRP-1 A, and sorLA, as estimated by surface plasmon resonance. The
binding of wild-type and variant complexes to the receptors was estimated by Biacore. Binding analyses were performed at several uPA–PAI-1 concentrations between 1.5 and 100 n
M
with the same receptor chips at two different days. The k
1
, k
)1
, and K
d
values were estimated for each separate binding curve from the initial rate of association and the rate of dissoci-
ation (as described in detail in Results and Experimental procedures). In some individual experiments, the association or dissociation time courses could not be analysed, and hence the K
d
values could not be calculated, because of irregular curve shapes caused by air bubbles, etc. Means, standard deviations, and numbers of determinations are indicated, the latter in paren-
theses.
uPA variant PAI-1 variant

VLDLR binding LRP-1 A binding SorLA binding
k
1
· 10
)5
(M
)1
Æs
)1
)
k
)1
· 10
4
(s
)1
)
K
d
(nM)
k
1
· 10
)5
(M
)1
Æs
)1
)
k

)1
· 10
4
(s
)1
)
K
d
(nM)
k
1
· 10
)5
(M
)1
Æs
)1
)
k
)1
· 10
4
(s
)1
)
K
d
(nM)
Wild-type Wild-type 2.73 ± 1.22
(10)

4.35 ± 1.22
(9)
1.60 ± 0.47
(8)
11.5 ± 6.0
(10)
4.59 ± 1.15
(4)
0.226 ± 0.054
(4)
7.05 ± 4.20
(10)
3.01 ± 1.37
(10)
0.480 ± 0.153
(10)
R178(35)A-R179
(36)A-R181(37a)A
Wild-type 2.62 ± 0.83
(10)
5.90 ± 1.17
(9)
a
2.35 ± 0.60
(9)
a
9.30 ± 3.86
(10)
5.27 ± 0.74
(3)

0.367 ± 0.051
(3)
c
6.53 ± 3.14
(10)
4.00 ± 0.99
(10)
0.704 ± 0.241
(10)
c
Wild-type K71A-R78A-
Y81A-K82A
2.03 ± 0.56
(9)
7.97 ± 1.96
(9)
a
4.07 ± 1.12
(8)
a
7.51 ± 2.54
(10)
10.3 ± 2.30
(4)
a
1.02 ± 0.26
(4)
a
6.73 ± 3.31
(10)

5.96 ± 0.53
(9)
a
1.02 ± 0.39
(9)
a
R178(35)A-R179
(36)A-R181(37a)A
K71A-R78A-
Y81A-K82A
0.494 ± 0.368
(9)
b
9.76 ± 3.63
(6)
27.2 ± 18.9
(6)
b
1.80 ± 1.43
(9)
b
27.3 ± 5.4
(4)
b
8.79 ± 2.69
(4)
b
3.12 ± 3.24
(9)
8.43 ± 1.00

(9)
b
5.66 ± 3.72
(9)
b
a
Significantly different from binding observed with wild-type uPA–wild-type PAI-1 complex (P<0.01).
b
Values for the complex with mutations in both uPA and PAI-1, which are significantly different
from the wild-type uPA–K71A-R78A-Y81A-K82A PAI-1 complex (P<0.01).
c
Significantly different from the wild-type uPA–wild-type PAI-1 complex (P<0.025).
Table 2. Effect of substitutions of PAI-1 residues on the accumula-
tion of uPA–PAI-1 on U937 cell surface. U937 cells were allowed
to bind the respective uPA–PAI-1 mutant complexes for 1 h on ice
and subsequently washed before incubation for 8 min at either 0 or
37 °C. Cell surface-associated complex was then released with a
low-pH buffer. Cell surface-associated and internalized complexes
are expressed as percentage of total amount of cell-associated
complex. Mean ± SD values are given for experiments performed
in triplicate.
PAI-1 variant
0 °C, 8 min 37 °C, 8 min
Inside
Cell surface
associated Inside
Cell surface
associated
Wild-type 5.7 ± 0.1 94.3 ± 0.1 39.2 ± 1.1 60.8 ± 1.1
H5A-P6A-Q109A 5.0 ± 0.2 95.0 ± 0.2 35.8 ± 1.8 64.2 ± 1.8

H79A-F116A-R117A 4.5 ± 0.2
a
95.5 ± 0.2
a
20.6 ± 0.2
a
79.4 ± 0.2
a
a
Significantly different from the corresponding number for wild-
type PAI-1 (P<0.01).
uPA–PAI-1 complex receptor binding S. Skeldal et al.
5150 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
the PAI-1 residues implicated in binding than of all
the residues tested. This finding agrees well with the
hypothesis that ligand recognition by this receptor
class relies on electrostatic interactions between basic
residues in the ligands and acidic residues in the recep-
tors [21]. However, X-ray crystal structure analysis of
LDLR at pH 5 [4] and receptor–ligand complexes
[2,3], site-directed mutagenesis of CTRs [30] (see
below) and our results presented here (Fig. 2) suggest
that hydrophobic interactions may also be important.
Almost all the residues implicated in binding were
localized in and around a-helix D and a-helix E, but,
interestingly, substitution of Arg270, localized in
b-strand 2C, % 2.7 nm from a-helix D, also affected
binding, suggesting an even more extended binding
surface (Fig. 7).
In uPA, we have shown that substitutions in and near

the 37-loop and 60-loop caused a substantial reduction
in binding. Thus, our findings for the first time implicate
the SPD of uPA in the binding of the complex to the
receptors. Obviously, we cannot exclude the possibility
that additional residues in the SPD contribute to the
binding. In addition, we could confirm the previously
implicated importance of the A-chain of uPA in the
binding of the uPA–PAI-1 complex to the receptors.
Moreover, we demonstrated that deletion of the uPA
growth factor domain resulted in reduced binding. As
the reduction was smaller than that caused by deletion
of the entire A-chain, it seems likely that both the
growth factor domain and the kringle contribute to the
binding, but putative endocytosis receptor-binding resi-
dues in the kringle remain unknown. The involvement
of the growth factor domain readily explains the
previously reported reduction of the affinity of the
uPA–serpin complexes for LRP-1A and SorLA in
the presence of the cellular receptor for uPA, uPAR
[11,12,14], as uPAR binds to the growth factor domain
of uPA [31]. Binding to uPAR would thus shield endo-
cytosis receptor-binding residues.
In the model of the complex between PAI-1 and the
uPA SPD, constructed from the 3D structure of the
a
1
-PI–trypsin complex, as determined by X-ray crystal
structure analysis [16], the residues of the 37-loop and
60-loop of uPA studied here are relatively close to
Fig. 6. Receptor-mediated degradation of uPA–PAI-1 wild-type and variant complexes in U937 and COS-1 cells. U937 cells were incubated

with 10 p
M
125
I-uPA–PAI-1 complex for 45 min at 37 °C, by which time the amount of degraded complex was determined as the fraction
of complex soluble in 7% trichloroacetic acid. The degradation of wild-type
125
I-uPA–PAI-1 complex was set equal to 1, and the degradation
of mutant complexes expressed relative to that. The figure shows mean ± SD for triple determinations in a typical experiment out of a
total of three with U937 cells and two with COS-1 cells. Insert,
125
I-RAP ligand blotting analysis of a membrane preparation from U937. Lane
1, purified VLDLR type II [44]; lane 2, membranes from U937 cells. The migration of molecular mass markers is indicated on the right.
S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5151
Fig. 7. The receptor binding surface in the uPA–PAI-1 complex. The uPA–PAI-1 complex shown is a SWISSPDBVIEWER surface display of a
model constructed from the structure of the a
1
-PI–trypsin complex [16] (pdb file 1EZX), by overlayering the a
1
-PI part of that structure with
the structure of cleaved PAI-1 [54] (pdb file 9PAI) and the trypsin part of that structure with the structure of the SPD of uPA [27] (pdb file
1LMW). Although the relative orientation of uPA and PAI-1 in the model is realistic, it does not allow any predictions of exact distances
between amino-acid residues. The effects of Ala substitutions on the binding of 20 p
M
125
I-uPA–PAI-1 complex to VLDLR, LRP-1A, and sor-
LA individually are depicted in the model. Ala substitution of residues colored red resulted in a more than fivefold reduction in binding. Ala
substitution of residues colored pink resulted in a 1.5–5-fold reduction in binding. PAI-1 residue Tyr81 and uPA residue Arg178(35), which
would have been colored pink, are not visible with the orientation of the structure used. Ala substitution of residues colored blue resulted in
a less than 1.5-fold reduction in binding when tested individually or less than twofold reduction when tested as part of triple mutants. One

structure is shown for binding to each receptor. In addition, the complex model is also shown as a ribbon diagram in which b-strand 2C of
PAI-1 is colored yellow and a-helix D and E of PAI-1 are colored green. The 37-loop and 60-loop of uPA are shown in red and pink, respect-
ively. Also indicated is a ribbon diagram of the pH 5 structure of the CTRs from LDLR [4] (pdb file 1 N7D).
uPA–PAI-1 complex receptor binding S. Skeldal et al.
5152 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
PAI-1, and it may be argued that the changed endocy-
tosis receptor binding following substitution of these
residues is due to a change in the relative orientations
of uPA and PAI-1. However, in the two available crys-
tal structure complexes [16,17], the corresponding resi-
dues remain fully surface-exposed. Secondly, the
substitution of these residues did not lead to changes
in the stability of the complex, as evaluated by anti-
body-binding assays performed in parallel with micro-
titer well receptor-binding assays (data not shown).
Thirdly, monoclonal antibodies with epitopes in the
37-loop and 60-loop remain accessible in the complex
[26]. Fourthly, in contrast with monoclonal antibodies
binding near a-helix F of PAI-1, a monoclonal anti-
body to PAI-1 with an epitope near the expected inter-
action phase with uPA did not induce lability of the
complex [32]. Fifthly, the absence of any large struc-
tural effects of the triple 37-loop uPA mutation was
established by demonstrating unchanged enzyme activ-
ity. The most obvious interpretation of the observed
effects of the substitutions of the residues in the 37
and 60 loops is that these residues are directly situated
at the surface of the interaction between the uPA–
PAI-1 complex and the endocytosis receptors.
In the model of the complex between PAI-1 and the

uPA SPD, constructed from the 3D structure of the
a
1
-PI–trypsin complex, as determined by X-ray crystal
structure analysis [16], the residues involved in binding
form a continuous, extended interaction surface span-
ning both PAI-1 and uPA (Fig. 7). Together with the
involvement of the uPA A-chain, these findings argue
for the possibility that the high affinity of the complex,
compared with that of the individual components, is
due to an extended interaction surface assembled by
complex formation. Our results do not exclude the
possibility that conformational changes or exposure of
‘cryptic’ sites on complex formation [13,23] contribute
to the increased affinity, but are more readily
explained by an avidity effect caused by two or more
binding sites being brought together on the same lig-
and. Thus, it should be noted that the 37-loop of uPA
does not seem to be in the part of uPA that changes
conformation on complex formation [33]. Another
argument against the ‘cryptic’ site theory is that the
3D structures of the regions of cleaved a
1
-PI and tryp-
sin-complexed a
1
-PI corresponding to the regions of
PAI-1 implicated in endocytosis receptor binding are
superimposable [16,34]. Assuming that cleaved and
complexed PAI-1 are as similar as cleaved and com-

plexed a
1
-PI, the observation of a K
d
for uPA–PAI-1
complex–LRP-1A binding of 0.4 nm,aK
d
for active
PAI-1–LRP-1A binding of 55 nm, and a K
d
for
cleaved PAI-1–LRP-1A binding of 41 n m [12] must
lead to the conclusion that conformational changes in
PAI-1 are unlikely to be the only cause of the
increased endocytosis receptor affinity upon complex
formation.
Ligand binding by receptors of the LDLR family is
critically dependent on the CTR clusters. VLDLR and
sorLA have only one such cluster, with eight or 11
CTRs, respectively. LRP-1A has four clusters, of two,
eight, 10, and 11 CTRs. In general, high affinity and
high specificity in ligand recognition by receptors of the
LDLR receptor family is believed to rely on several indi-
vidual CTRs making contact with several receptor
recognition patches on the same ligand. Each of the 217
known human CTRs comprises %40 amino acids. A
conserved scaffold motif of only 12 amino acids is
required for this common structure, with 28 noncon-
served positions. Receptors from the LDLR family
alone accounts for 132 different CTRs. Crystallographic

and NMR studies of individual repeats have revealed
that the sequence variability in short loop regions of
each repeat results in a unique contour surface and
charge density for each repeat [35]. Several observations
suggest that the binding sites for plasminogen activator–
PAI-1 complexes are also assembled from combinations
of CTRs. Thus, although the interpretation of our pre-
sent results with LRP-1A is hampered by the fact that
both its CTR cluster II and IV bind to the uPA–PAI-1
complex [36–39], the results do demonstrate that the
uPA–PAI-1 complex has slightly different interaction
surfaces with each of the three receptors tested (Fig. 7).
This observation is in agreement with conclusions from
previous studies that VLDLR and LRP-1A have over-
lapping but not identical binding specificities in a series
of serine protease–serpin complexes [40]. By expression
of VLDLR variants lacking specific complement-type
repeats, it was found that the second CTR in the cluster
of eight in this receptor was required for maximal affin-
ity for the uPA–PAI-1 complex [41]. The tPA–PAI-1
complex and PAI-1 were reported to bind to a fragment
encompassing CTR3-CTR4-CTR5-CTR6-CTR7 of
CTR cluster II of LRP-1A (the CTRs of the most N-ter-
minal CTR cluster being numbered 1 and 2) [38,42,43].
The minimal functional unit capable of binding the
uPA–PAI-1 complex to LRP-1A CTR cluster II was
reported to be a two-CTR fragment, and the CTR5–
CTR6 pair to be the pair mainly responsible for binding
the uPA–PAI-1 complex [30]. From the dimensions of
the CTRs emerging from the X-ray crystal structure

analysis of LDLR [4], the uPA SPD and the PAI-1
region around a-helix D and a-helix E would each make
contact with a CTR. To also accommodate the A-chain
of uPA and the more distantly localized R270 of PAI-1,
two additional CTRs would be needed. Thus, whereas
S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5153
the individual components uPA and PAI-1 would each
only be able to make contact with one or two CTRs, we
propose that the high affinity of the complex is related
to contacts with three or four CTRs.
In conclusion, we here add several new details to the
map of the endocytosis receptor-binding surface of ser-
ine protease–serpin complexes and for the first time
implicate the SPD of uPA in the binding. The residues
that mediate the binding affinity of the A-chain of uPA
remain to be established. In the long term, exact
mapping of the interaction surface will require X-ray
structure analysis of cocrystals between the uPA–PAI-1
complex and one of the receptors.
Experimental procedures
Cell lines
U937 cells were cultured in RPMI 1640 medium. COS-1
and HEK293T cells were cultured in Dulbecco modified
Eagle’s medium. Both media were supplemented with 10%
fetal bovine serum, 100 UÆmL
)1
penicillin and 100 lgÆmL
)1
streptomycin. The medium for HEK293T cells also con-

tained 1% nonessential amino acids. Cells were maintained
in a humidified atmosphere with 5% CO
2
at 37 °C. CHO
cells were cultured as described [44].
Receptors
A stop codon succeeding the codon encoding amino acid
772 was introduced into human VLDLR type-I cDNA, i.e.
cDNA containing all exons of the human VLDLR gene
(kindly provided by A. Soutar, University College London,
UK) [44]. The resulting cDNA, encoding soluble VLDLR-I
without the cytoplasmic domain and transmembrane region
(residues 1–772, sVLDLR), was cloned into the plasmid
pcDNA3.1- and controlled by DNA sequencing. Transient
transfection by the calcium phosphate precipitation method
and expression of soluble VLDLR were carried out by
standard methods [26]. Briefly, HEK293T cells were grown
to full confluence and split 1 : 2 and allowed to grow for
an additional 24 h. Cells were then split 1 : 3 and seeded
on gelatine-coated dishes. After 24 h, chloroquine was
added directly to the cultures to a final concentration of
25 lm. The cultures were then incubated for 1 h. Then
30 lg DNA and water were mixed to a final volume of
1752 lL, and 248 lL2m CaCl
2
was added. This mixture
was added dropwise to 2 mL 42 mm Hepes, pH 7.05, con-
taining 274 mm NaCl, 10 mm KCl, 1.5 mm Na
2
HPO

4
, and
11 mmd-(+)-glucose. This solution was then added drop-
wise to the conditioned medium of one 15-cm culture dish
with HEK293T cells. After 7–11 h of incubation, the med-
ium was replaced with serum-free medium, which was har-
vested 48 h later. This was repeated twice. The conditioned
medium containing soluble VLDLR was poured on to an
RAP-coupled Sepharose column, recirculated for 24 h, and
washed. The equilibration and washing buffer was Hepes-
buffered saline (10 mm Hepes, pH 7.4, 0.14 m NaCl) with
2mm CaCl
2
and 1 mm MgCl
2
. Elution was carried out
with 0.1 m acetic acid (pH 3.5) ⁄ 0.5 m NaCl ⁄ 10 mm EDTA.
Fractions were neutralized immediately with 0.1 vol. 1 m
Tris (pH 9) ⁄ 25 mm CaCl
2
. Relevant fractions were selected
on the basis of
125
I-RAP-blots, and pooled. The receptor
preparations were concentrated with a Centriplus YM-10
spin column, dialysed into NaCl ⁄ P
i
(10 mm sodium phos-
phate, pH 7.4, 0.14 m NaCl), and stored at )80 °C.
Full-length LRP-1A was purified from human placentas

as described [45]. Mature, truncated soluble human sorLA
(residues 54–2107), lacking the N-terminal propeptide and
the C-terminal cytoplasmic and transmembrane part, was
produced in CHO cells and purified as described [46].
Proteases, inhibitors, and their complexes
Human wild-type uPA and uPA mutants were expressed
recombinantly in and purified from conditioned media of
HEK293T cells, as previously described [26]. Recombinant
uPA N-terminally truncated at Lys136 (LMW-uPA) was a
gift from W Gu
¨
nzler, Gru
¨
nenthal Company, Aachen, Ger-
many. uPA variants were characterized with respect to K
m
for hydrolysis of S-2444 (pyro-Glu-Gly-Arg-p-nitroanilide)
as previously described [47].
Human wild-type PAI-1 and PAI-1 mutants were
expressed recombinantly in and purified from conditioned
media of HEK293T cells, as previously described [24].
Some PAI-1 variants were characterized with respect to spe-
cific inhibitory activity, second-order rate constant for the
reaction with uPA, and relative vitronectin affinity, as des-
cribed previously [26,48]. The numbering of PAI-1 residues
was as described previously [49].
125
I-labeling of uPA, LMW-uPA, and RAP was per-
formed as described [12]. Complexes between
125

I-uPA and
purified PAI-1 were prepared as described [40]. In some
cases, complexes between wild-type
125
I-uPA and PAI-1
variants were prepared by the use of conditioned media
from cells transfected with the corresponding PAI-1 cDNA.
In these cases, the conditioned media were incubated for
1 h at room temperature with 0.1% SDS, followed by an
eightfold dilution with ice-cold 0.114 m Tris ⁄ HCl, pH 7.4,
containing 1.14% BSA, in order to re-activate the latent
PAI-1.
125
I-uPA was then added in an amount correspond-
ing to a 15-fold molar excess of PAI-1 over
125
I-uPA. In
experiments performed to compare different PAI-1 variants,
aliquots of one and the same preparation of
125
I-uPA were
added to separate portions of conditioned medium with the
different PAI-1 variants. The
125
I-uPA–PAI-1 complexes
were allowed to form during 30 min of incubation at
37 °C. The
125
I-uPA–PAI-1 complexes were purified by
pouring the uPA–PAI-1 mixture on to an anti-uPA

uPA–PAI-1 complex receptor binding S. Skeldal et al.
5154 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
mAb6-coupled Sepharose column followed by an anti-PAI-
1 mAb2-coupled Sepharose column. Both columns were
equilibrated and washed in washing buffer (0.1 m Tris, 1 m
NaCl, 1% BSA) and eluted with 0.1 m acetic acid
(pH 2.9) ⁄ 1 m NaCl, and fractions immediately neutralized
with 0.1 vol. 1 m Tris (pH 9) ⁄ 1 m NaCl.
125
I-LMW-uPA–
PAI-1 complex was produced and purified in the same way
as uPA–PAI-1 except that an anti-uPA mAb2-coupled
Sepharose column was used instead of an anti-uPA mAb6-
coupled Sepharose column.
To prepare nonradioactive wild-type and mutant uPA–
PAI-1 complex, recombinant PAI-1 expressed in and puri-
fied from HEK293T cells in the latent form (see above)
were denatured and refolded by the use of guanidinium
chloride [24]. The re-activated PAI-1 was mixed with puri-
fied uPA also expressed in and purified from HEK293T
cells (see above). The complexes formed were purified by
immuno-affinity chromatography on two successive col-
umns coupled with monoclonal antibodies against PAI-1
and uPA, respectively [12].
Antibodies and other proteins
Anti-PAI-1 mAb2, anti-uPA mAb2, anti-uPA mAb6, and
anti-uPA mAb3689 have previously been described [26,50].
Recombinant RAP was a gift from M Etzerodt, Depart-
ment of Molecular Biology, University of Aarhus [5].
Solid-phase binding assays

Receptors were coated on to microtiter wells overnight in
50 mm NaHCO
3
, pH 9.6, with a receptor density adjusted to
obtain 5–10% binding of wild-type
125
I-uPA–PAI-1 com-
plex, corresponding to receptor concentrations in the coating
buffer of % 1 lgÆmL
)1
. The wells were blocked with 10%
BSA in binding buffer (10 mm Hepes, pH 7.8, 140 mm NaCl,
2mm CaCl
2
,1mm MgCl
2
) for 2 h, washed, and incubated
overnight at 4 °C with 20 pm
125
I-ligand. Washing and incu-
bation with ligands were performed in binding buffer with
1% BSA. When the effect of monoclonal antibodies on
binding was investigated, the antibodies were added at a con-
centration of 20 lgÆmL
)1
. After the incubation, the super-
nantants were removed for c-counting and the wells washed.
Bound ligand was released from the wells with 10% SDS and
taken for c-counting. Ligand binding was expressed as
bound divided by free ligand and corrected for nonspecific

binding, i.e. the radioactivity recovered from wells without
receptor. As a control for the integrity of the complexes, the
binding of all Ala-substituted complexes to parallel wells
coated with 20 lgÆmL
)1
anti-PAI-1 mAb2 was shown not to
differ more than 20% from that of the wild-type. The bind-
ing of variant uPA–PAI-1 complexes to the receptors was
corrected by the use of the small variations in antibody bind-
ing, assumed to be caused by a dissociation of small fractions
of the complexes, by dividing the actually observed binding
by the factor by which antibody binding of each variant
complex differed from that of the wild-type complex.
Ligand degradation and internalization
experiments
For ligand degradation experiments with U937 cells, the
cells were washed three times with serum-free medium sup-
plemented with 0.5% BSA and incubated for 45 min at
37 °C in 0.5 mL of the same buffer with 10 pm
125
I-labeled
ligand and a cell density of 10 · 10
6
cellsÆmL
)1
. The
cultures were then made 7% with respect to trichloroacetic
acid and centrifuged. Supernatants and pellets were c-coun-
ted separately. For experiments with COS-1 cells, the cells
were cultured in 24-well cell culture dishes to a confluence

% 80%, washed three times with serum-free medium supple-
mented with 0.5% BSA, and incubated for 4 h at 37 °C
with 0.3 mL of the same buffer with 10 pm
125
I-labeled lig-
and. The medium was then removed, and the cells solubi-
lized with 1% SDS. BSA was added to the cell extracts to a
final concentration of 1%. Cell medium and cell extracts
were then made 7% with respect to trichloroacetic acid and
centrifuged. Supernatant and pellets were c-counted sepa-
rately. Ligand degradation was defined as nontrichloroace-
tic acid-precipitable radioactivity versus total added
radioactivity. The degradation in incubations with 100 nm
RAP was scored as nonspecific degradation, not mediated
by RAP-binding receptors, and subtracted from all other
data [51]. Separate control experiments showed that there
was linearity between time and amount of ligand degraded
with the incubation times used here (not shown).
To evaluate the distribution of
125
I-uPA–PAI-1 com-
plexes between the cell surface and intracellular vesicles, we
washed U937 cells three times with serum-free medium con-
taining 0.5% BSA and cell density adjusted to 10 · 10
6
cell-
sÆmL
)1
in the same buffer. Then 0.5 mL cell suspension was
incubated with 10 pm

125
I-uPA–PAI-1 for 1 h on ice with
gentle rocking and then washed three times in ice-cold
washing buffer. Cells where then incubated at 37 °C or kept
on ice for 8 min and subsequently quickly placed on ice.
Cell surface-associated ligand was released by washing cells
three times with an ice-cold buffer comprising 0.2 m acetic
acid, pH 2.6, and 0.1 m NaCl. The remaining radioactivity
was defined as internalized ligand. Internalized and cell sur-
face-associated ligand were expressed as a percentage of the
total radioactivity.
125
I-RAP blotting analysis of membrane
fragments from U937 cells
Membrane fragments from U937 cells were purified as
described previously [44]. Ligand blotting analysis with
125
I-RAP was carried out as previously described [52],
using VLDLR type-II as a molecular mass marker [44].
S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5155
Surface plasmon resonance analysis
Surface plasmon resonance measurements were performed
on a Biacore 3000 instrument (Biacore, Uppsala, Sweden)
equipped with CM5 sensor chips. The receptors were
immobilized to densities of % 50 fmolÆmm
)2
, and samples
for binding (40 lL) were injected at 5 lLÆmin
)1

at 25 °Cin
10 mm Hepes, pH 7.4, containing 150 mm NaCl, 1.5 mm
CaCl
2
,1mm EGTA, and 0.01% P20. Various concentra-
tions of uPA–PAI-1 complex were injected for 8 min. After
discontinuation of the injection, dissociation of bound lig-
and was followed for another 8 min. Binding was expressed
in relative response units as the response obtained from the
flow cell containing immobilized receptor minus the
response obtained when using an activated but uncoupled
chip.
The K
d
values were calculated as the ratio between the
dissocation rate constants (k
)1
values) and the association
rate constants (k
1
values). The dissocation rate constants
were estimated from semilogarithmic plots of the relative
amount of uPA–PAI-1 complex bound to the chip versus
time during the dissociation phase. The data were fitted to
biphasic dissociation kinetics, to distinguish between slowly
dissociating, high-affinity binding (rate constant k
)1
), and
low affinity, rapidly dissociating binding (rate constant
k

)2
):
Relative binding ¼ A expðÀk
À1
tÞþB expðÀk
À2
tÞð1Þ
To estimate the association rate constants (k
1
values),
we estimated the initial rates of association by linear
regression analysis. The initial rates were converted
into k
1
values by the following equation:
Initial rate ¼ k
1
½R
T
½uPA À PAI-1 complexð2Þ
Here, the [R
T
] values were calculated from the response units
bound to the chips when exposed to 100 nm wild-type uPA–
wild-type PAI-1 complex, after correction for the rapidly dis-
sociating fraction of % 20%. In all cases, the initial associ-
ation rates showed the expected relationship with the uPA–
PAI-1 complex concentration. The K
d
values were then cal-

culated as k
)1
⁄ k
1
.
Statistical analysis
Results were analysed by Student’s t -test.
Acknowledgements
We acknowledge the Danish Cancer Research Founda-
tion, The Interdisciplinary Nanoscience Centre of Aar-
hus University (iNANO), the Danish Cancer Society,
the Danish National Research Foundation, and Novo
Nordisk Foundation.
References
1 Bu G & Schwartz AL (1998) RAP, a novel type of ER
chaperone. Trends Cell Biol 8, 272–276.
2 Fisher C, Beglova N & Blacklow SC (2006) Structure of
an LDLR-RAP complex reveals a general mode for
ligand recognition by lipoprotein receptors. Mol Cell 22,
277–283.
3 Verdaguer N, Fita I, Reithmayer M, Moser R & Blaas
D (2004) X-ray structure of a minor group human rhi-
novirus bound to a fragment of its cellular receptor pro-
tein. Nat Struct Mol Biol 11, 429–434.
4 Rudenko G, Henry L, Henderson K, Ichtchenko K,
Brown MS, Goldstein JL & Deisenhofer J (2002) Struc-
ture of the LDL receptor extracellular domain at endo-
somal pH. Science 298, 2353–2358.
5 Nykjaer A, Petersen CM, Moller B, Jensen PH,
Moestrup SK, Holtet TL, Etzerodt M, Thogersen

HC, Munch M, Andreasen PA, et al. (1992) Purified
alpha 2-macroglobulin receptor ⁄ LDL receptor-related
protein binds urokinase.plasminogen activator inhib-
itor type-1 complex. Evidence that the alpha 2-macro-
globulin receptor mediates cellular degradation of
urokinase receptor-bound complexes. J Biol Chem
267, 14543–14546.
6 Moestrup SK, Nielsen S, Andreasen P, Jorgensen KE,
Nykjaer A, Roigaard H, Gliemann J & Christensen EI
(1993) Epithelial glycoprotein-330 mediates endocytosis
of plasminogen activator-plasminogen activator inhibi-
tor type-1 complexes. J Biol Chem 268, 16564–16570.
7 Willnow TE, Goldstein JL, Orth K, Brown MS &
Herz J (1992) Low density lipoprotein receptor-related
protein and gp330 bind similar ligands, including plas-
minogen activator-inhibitor complexes and lactoferrin,
an inhibitor of chylomicron remnant clearance. J Biol
Chem 267, 26172–26180.
8 Heegaard CW, Simonsen AC, Oka K, Kjoller L, Chris-
tensen A, Madsen B, Ellgaard L, Chan L & Andreasen
PA (1995) Very low density lipoprotein receptor binds
and mediates endocytosis of urokinase-type plasminogen
activator-type-1 plasminogen activator inhibitor com-
plex. J Biol Chem 270, 20855–20861.
9 Argraves KM, Battey FD, MacCalman CD, McCrae KR,
Gafvels M, Kozarsky KF, Chappell DA, Strauss JF, 3rd
& Strickland DK (1995) The very low density lipoprotein
receptor mediates the cellular catabolism of lipoprotein
lipase and urokinase-plasminogen activator inhibitor type
I complexes. J Biol Chem 270, 26550–26557.

10 Li Y, Knisely JM, Lu W, McCormick LM, Wang J,
Henkin J, Schwartz AL & Bu G (2002) Low density
lipoprotein (LDL) receptor-related protein 1B impairs
urokinase receptor regeneration on the cell surface
and inhibits cell migration. J Biol Chem 277, 42366–
42371.
uPA–PAI-1 complex receptor binding S. Skeldal et al.
5156 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
11 Gliemann J, Hermey G, Nykjaer A, Petersen CM,
Jacobsen C & Andreasen PA (2004) The mosaic recep-
tor sorLA ⁄ LR11 binds components of the plasminogen-
activating system and platelet-derived growth factor-BB
similarly to LRP1 (low-density lipoprotein receptor-rela-
ted protein), but mediates slow internalization of bound
ligand. Biochem J 381, 203–212.
12 Nykjaer A, Kjoller L, Cohen RL, Lawrence DA, Garni-
Wagner BA, Todd RF, 3rd van Zonneveld AJ, Gli-
emann J & Andreasen PA (1994) Regions involved in
binding of urokinase-type-1 inhibitor complex and pro-
urokinase to the endocytic alpha 2-macroglobulin recep-
tor ⁄ low density lipoprotein receptor-related protein.
Evidence that the urokinase receptor protects pro-uro-
kinase against binding to the endocytic receptor. J Biol
Chem 269, 25668–25676.
13 Horn IR, van den Berg BM, Moestrup SK, Pannekoek
H & van Zonneveld AJ (1998) Plasminogen activator
inhibitor 1 contains a cryptic high affinity receptor bind-
ing site that is exposed upon complex formation with
tissue-type plasminogen activator. Thromb Haemost 80,
822–828.

14 Croucher D, Saunders DN & Ranson M (2006) The
urokinase ⁄ PAI-2 complex: a new high affinity ligand for
the endocytosis receptor low density lipoprotein recep-
tor-related protein. J Biol Chem 281, 10206–10213.
15 Ye, S & Goldsmith EJ (2001) Serpins and other covalent
protease inhibitors. Curr Opin Struct Biol 11, 740–745.
16 Huntington JA, Read RJ & Carrell RW (2000) Struc-
ture of a serpin-protease complex shows inhibition by
deformation. Nature 407, 923–926.
17 Dementiev A, Dobo J & Gettins PG (2006) Active site
distortion is sufficient for proteinase inhibition by ser-
pins: structure of the covalent complex of alpha1-protei-
nase inhibitor with porcine pancreatic elastase. J Biol
Chem 281, 3452–3457.
18 YeS, Cech AL, Belmares R, Bergstrom RC, Tong Y,
Corey DR, Kanost MR & Goldsmith EJ (2001) The
structure of a Michaelis serpin-protease complex. Nat
Struct Biol 8, 979–983.
19 Dementiev A, Simonovic M, Volz K & Gettins PG
(2003) Canonical inhibitor-like interactions explain reac-
tivity of alpha1-proteinase inhibitor Pittsburgh and
antithrombin with proteinases. J Biol Chem 278, 37881–
37887.
20 Dementiev A, Petitou M, Herbert JM & Gettins PG
(2004) The ternary complex of antithrombin-anhydro-
thrombin-heparin reveals the basis of inhibitor specifi-
city. Nat Struct Mol Biol 11, 863–867.
21 Andreasen PA, Sottrup-Jensen L, Kjoller L, Nykjaer
A, Moestrup SK, Petersen CM & Gliemann J (1994)
Receptor-mediated endocytosis of plasminogen activa-

tors and activator ⁄ inhibitor complexes. FEBS Lett
338, 239–245.
22 Rodenburg KW, Kjoller L, Petersen HH & Andreasen
PA (1998) Binding of urokinase-type plasminogen acti-
vator-plasminogen activator inhibitor-1 complex to the
endocytosis receptors alpha2-macroglobulin receptor ⁄
low-density lipoprotein receptor-related protein and
very-low-density lipoprotein receptor involves basic resi-
dues in the inhibitor. Biochem J 329, 55–63.
23 Stefansson S, Muhammad S, Cheng XF, Battey FD,
Strickland DK & Lawrence DA (1998) Plasminogen
activator inhibitor-1 contains a cryptic high affinity
binding site for the low density lipoprotein receptor-
related protein. J Biol Chem 273, 6358–6366.
24 Munch M, Heegaard CW & Andreasen PA (1993)
Interconversions between active, inert and substrate
forms of denatured ⁄ refolded type-1 plasminogen activa-
tor inhibitor. Biochim Biophys Acta 1202, 29–37.
25 Andreasen PA, Kjoller L, Christensen L & Duffy MJ
(1997) The urokinase-type plasminogen activator system
in cancer metastasis: a review. Int J Cancer 72, 1–22.
26 Petersen HH, Hansen M, Schousboe SL & Andreasen
PA (2001) Localization of epitopes for monoclonal anti-
bodies to urokinase-type plasminogen activator: rela-
tionship between epitope localization and effects of
antibodies on molecular interactions of the enzyme. Eur
J Biochem 268, 4430–4439.
27 Spraggon G, Phillips C, Nowak UK, Ponting CP, Saun-
ders D, Dobson CM, Stuart DI & Jones EY (1995) The
crystal structure of the catalytic domain of human uro-

kinase-type plasminogen activator. Structure 3, 681–691.
28 Martensen PM, Oka K, Christensen L, Rettenberger
PM, Petersen HH, Christensen A, Chan L, Heegaard
CW & Andreasen PA (1997) Breast carcinoma epithelial
cells express a very low-density lipoprotein receptor var-
iant lacking the O-linked glycosylation domain encoded
by exon 16, but with full binding activity for serine
proteinase ⁄ serpin complexes and Mr-40,000 receptor-
associated protein. Eur J Biochem 248, 583–591.
29 Kjoller L, Simonsen AC, Ellgaard L & Andreasen PA
(1995) Differential regulation of urokinase-type-1 inhibi-
tor complex endocytosis by phorbol esters in different
cell lines is associated with differential regulation of
alpha 2-macroglobulin receptor and urokinase receptor
expression. Mol Cell Endocrinol 109, 209–217.
30 Andersen OM, Petersen HH, Jacobsen C, Moestrup
SK, Etzerodt M, Andreasen PA & Thogersen HC
(2001) Analysis of a two-domain binding site for the
urokinase-type plasminogen activator-plasminogen acti-
vator inhibitor-1 complex in low-density-lipoprotein-
receptor-related protein. Biochem J 357, 289–296.
31 Ploug M, Gardsvoll H, Jorgensen TJ, Lonborg Hansen
L & Dano K (2002) Structural analysis of the interac-
tion between urokinase-type plasminogen activator and
its receptor: a potential target for anti-invasive cancer
therapy. Biochem Soc Trans 30, 177–183.
S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5157
32 De Taeye B, Verbeke K, Compernolle G, Biesemans W,
Gils A & Declerck PJ (2003) Structural determinants in

the stability of the serpin ⁄ proteinase complex. Biochem
Biophys Res Commun 307, 529–534.
33 Egelund R, Petersen TE & Andreasen PA (2001) A ser-
pin-induced extensive proteolytic susceptibility of uroki-
nase-type plasminogen activator implicates distortion of
the proteinase substrate-binding pocket and oxyanion
hole in the serpin inhibitory mechanism. Eur J Biochem
268, 673–685.
34 Huntington JA, Pannu NS, Hazes B, Read RJ, Lomas
DA & Carrell RW (1999) A 2.6 A
˚
structure of a serpin
polymer and implications for conformational disease.
J Mol Biol 293, 449–455.
35 Silverman J, Liu Q, Bakker A, To W, Duguay A, Alba
BM, Smith R, Rivas A, Li P, Le H, et al. (2005) Multi-
valent avimer proteins evolved by exon shuffling of a
family of human receptor domains. Nat Biotechnol 23,
1556–1561.
36 Moestrup SK, Holtet TL, Etzerodt M, Thogersen HC,
Nykjaer A, Andreasen PA, Rasmussen HH, Sottrup-
Jensen L & Gliemann J (1993) Alpha 2-macroglobulin-
proteinase complexes, plasminogen activator inhibitor
type-1-plasminogen activator complexes, and receptor-
associated protein bind to a region of the alpha 2-
macroglobulin receptor containing a cluster of eight
complement-type repeats. J Biol Chem 268, 13691–
13696.
37 Willnow TE, Orth K & Herz J (1994) Molecular dissec-
tion of ligand binding sites on the low density lipopro-

tein receptor-related protein. J Biol Chem 269, 15827–
15832.
38 Neels JG, van Den Berg BM, Lookene A, Olivecrona
G, Pannekoek H & van Zonneveld AJ (1999) The sec-
ond and fourth cluster of class A cysteine-rich repeats
of the low density lipoprotein receptor-related protein
share ligand-binding properties. J Biol Chem 274,
31305–31311.
39 Obermoeller-McCormick LM, Li Y, Osaka H, Fitz-
Gerald DJ, Schwartz AL & Bu G (2001) Dissection of
receptor folding and ligand-binding property with func-
tional minireceptors of LDL receptor-related protein.
J Cell Sci 114, 899–908.
40 Kasza A, Petersen HH, Heegaard CW, Oka K, Chris-
tensen A, Dubin A, Chan L & Andreasen PA (1997)
Specificity of serine proteinase ⁄ serpin complex binding
to very-low-density lipoprotein receptor and alpha2-
macroglobulin receptor ⁄ low-density-lipoprotein-recep-
tor-related protein. Eur J Biochem 248, 270–281.
41 Mikhailenko I, Considine W, Argraves KM, Loukinov
D, Hyman BT & Strickland DK (1999) Functional
domains of the very low density lipoprotein receptor:
molecular analysis of ligand binding and acid-dependent
ligand dissociation mechanisms. J Cell Sci 112, 3269–
3281.
42 Horn IR, van den Berg BM, van der Meijden PZ,
Pannekoek H & van Zonneveld AJ (1997) Molecular
analysis of ligand binding to the second cluster of
complement-type repeats of the low density lipoprotein
receptor-related protein. Evidence for an allosteric

component in receptor-associated protein-mediated
inhibition of ligand binding. J Biol Chem 272, 13608–
13613.
43 Vash B, Phung N, Zein S & DeCamp D (1998) Three
complement-type repeats of the low-density lipoprotein
receptor-related protein define a common binding site
for RAP, PAI-1, and lactoferrin. Blood 92, 3277–3285.
44 Rettenberger PM, Oka K, Ellgaard L, Petersen HH,
Christensen A, Martensen PM, Monard D, Etzerodt M,
Chan L & Andreasen PA (1999) Ligand binding proper-
ties of the very low density lipoprotein receptor.
Absence of the third complement-type repeat encoded
by exon 4 is associated with reduced binding of Mr
40,000 receptor-associated protein. J Biol Chem 274,
8973–8980.
45 Moestrup SK & Gliemann J (1991) Analysis of ligand
recognition by the purified alpha 2-macroglobulin recep-
tor (low density lipoprotein receptor-related protein).
Evidence that high affinity of alpha 2-macroglobulin-
proteinase complex is achieved by binding to adjacent
receptors. J Biol Chem 266, 14011–14017.
46 Jacobsen L, Madsen P, Jacobsen C, Nielsen MS,
Gliemann J & Petersen CM (2001) Activation and func-
tional characterization of the mosaic receptor Sor-
LA ⁄ LR11. J Biol Chem 276, 22788–22796.
47 Hansen M, Wind T, Blouse GE, Christensen A, Peter-
sen HH, Kjelgaard S, Mathiasen L, Holtet TL &
Andreasen PA (2005) A urokinase-type plasminogen
activator-inhibiting cyclic peptide with an unusual P2
residue and an extended protease binding surface

demonstrates new modalities for enzyme inhibition.
J Biol Chem 280, 38424–38437.
48 Einholm AP, Pedersen KE, Wind T, Kulig P, Overg-
aard MT, Jensen JK, Bodker JS, Christensen A, Charl-
ton P & Andreasen PA (2003) Biochemical mechanism
of action of a diketopiperazine inactivator of plasmino-
gen activator inhibitor-1. Biochem J 373, 723–732.
49 Andreasen PA, Riccio A, Welinder KG, Douglas R,
Sartorio R, Nielsen LS, Oppenheimer C, Blasi F &
Dano K (1986) Plasminogen activator inhibitor type-1:
reactive center and amino-terminal heterogeneity deter-
mined by protein and cDNA sequencing. FEBS Lett
209, 213–218.
50 Wind T, Jensen MA & Andreasen PA (2001) Epitope
mapping for four monoclonal antibodies against human
plasminogen activator inhibitor type-1: implications for
antibody-mediated PAI-1-neutralization and vitronectin-
binding. Eur J Biochem 268, 1095–1106.
51 Jensen PH, Christensen EI, Ebbesen P, Gliemann J &
Andreasen PA (1990) Lysosomal degradation of recep-
uPA–PAI-1 complex receptor binding S. Skeldal et al.
5158 FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS
tor-bound urokinase-type plasminogen activator is
enhanced by its inhibitors in human trophoblastic chor-
iocarcinoma cells. Cell Regul 1, 1043–1056.
52 Simonsen AC, Heegaard CW, Rasmussen LK, Ellgaard
L, Kjoller L, Christensen A, Etzerodt M & Andreasen
PA (1994) Very low density lipoprotein receptor from
mammary gland and mammary epithelial cell lines binds
and mediates endocytosis of M (r) 40,000 receptor asso-

ciated protein. FEBS Lett 354, 279–283.
53 Sharp AM, Stein PE, Pannu NS, Carrell RW, Berken-
pas MB, Ginsburg D, Lawrence DA & Read RJ (1999)
The active conformation of plasminogen activator inhi-
bitor 1, a target for drugs to control fibrinolysis and cell
adhesion. Structure 7, 111–118.
54 Aertgeerts K, De Bondt HL, De Ranter CJ & Declerck
PJ (1995) Mechanisms contributing to the conforma-
tional and functional flexibility of plasminogen activator
inhibitor-1. Nat Struct Biol 2, 891–897.
S. Skeldal et al. uPA–PAI-1 complex receptor binding
FEBS Journal 273 (2006) 5143–5159 ª 2006 The Authors Journal compilation ª 2006 FEBS 5159