Báo cáo khoa học: Limited mutagenesis increases the stability of human carboxypeptidase U (TAFIa) and demonstrates the importance of CPU stability over proCPU concentration in down-regulating fibrinolysis doc

Limited mutagenesis increases the stability of human
carboxypeptidase U (TAFIa) and demonstrates the
importance of CPU stability over proCPU concentration
in down-regulating fibrinolysis
Wolfgang Knecht
1
, Johan Willemse
3
, Hanna Stenhamre
1
, Mats Andersson
2
, Pia Berntsson
1
,
Christina Furebring
2
, Anna Harrysson
1
, Ann-Christin Malmborg Hager
2
, Britt-Marie Wissing
1
,
Dirk Hendriks
3
and Philippe Cronet
1
1 AstraZeneca R & D Mo
¨
lndal, Mo

¨
lndal, Sweden
2 Alligator Bioscience AB, Lund, Sweden
3 Laboratory of Medical Biochemistry, University of Antwerp, Wilrijk, Belgium
The fragile balance between the activities of the coagu-
lation cascade (thrombin generation) and the fibrino-
lytic system (plasmin generation) is essential to prevent
excessive blood loss upon damage of a blood vessel,
while maintaining the blood flow in parts of the
body distant from the injury. Procarboxypeptidase U
[proCPU, thrombin-activatable fibrinolysis inhibitor
(TAFI), EC 3.4.17.20, MEROPS M14.009] belongs
to the metallocarboxypeptidase family and is a
human plasma zymogen, which is also known as
thrombin-activatable fibrinolysis inhibitor (TAFI),
plasma procarboxypeptidase B and procarboxypepti-
dase R [1,2]. ProCPU has been proposed to be a
molecular link between coagulation and fibrinolysis
[3,4]. The physiological role of proCPU and its activa-
ted form, carboxypeptidase U (CPU) is outlined in
Fig. 1. ProCPU is synthesized in the liver and secreted
into the plasma following the removal of its signal
Keywords
carboxypeptidase; coagulation; directed
evolution; fibrinolysis; protease
Correspondence
W. Knecht, Molecular Pharmacology –
Target Production, AstraZeneca R & D
Mo
¨

lndal, 431 83 Mo
¨
lndal, Sweden
Fax: + 46 317763753
Tel: + 46 317065341
E-mail:
(Received 5 November 2005, accepted
19 December 2005)
doi:10.1111/j.1742-4658.2006.05110.x
Procarboxypeptidase U [proCPU, thrombin-activatable fibrinolysis inhib-
itor (TAFI), EC 3.4.17.20] belongs to the metallocarboxypeptidase family
and is a zymogen found in human plasma. ProCPU has been proposed to
be a molecular link between coagulation and fibrinolysis. Upon activation
of proCPU, the active enzyme (CPU) rapidly becomes inactive due to its
intrinsic instability. The inherent instability of CPU is likely to be of major
importance for the in vivo down-regulation of its activity, but the under-
lying structural mechanisms of this fast and spontaneous loss of activity of
CPU have not yet been explained, and they severely inhibit the structural
characterization of CPU. In this study, we screened for more thermostable
versions of CPU to increase our understanding of the mechanism underly-
ing the instability of CPU’s activity. We have shown that single as well as
a few 2–4 mutations in human CPU can prolong the half-life of CPU’s
activity at 37 °C from 0.2 h of wild-type CPU to 0.5–5.5 h for the mutants.
We provide evidence that the gain in stable activity is accompanied by a
gain in thermostability of the enzyme and increased resistance to proteo-
lytic digest by trypsin. Using one of the stable mutants, we demonstrate
the importance of CPU stability over proCPU concentration in down-regu-
lating fibrinolysis.
Abbreviations
BEVS, Baculovirus expression vector system; CLT, clot lysis time; CPB, carboxypeptidase B; CPU, carboxypeptidase U; EPP, error prone

PCR; Hip-Arg, hippuryl-
L-arginine; ORF, open reading frame; PTCI, potato tuber carboxypeptidase inhibitor; TAFI, thrombin-activatable
fibrinolysis inhibitor; WT, wild type.
778 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS
peptide (prepeptide, see Fig. 2). It can be activated
from its zymogen form to CPU by thrombin, plasmin
or most efficiently the thrombin–thrombomodulin
complex by cleavage after R114 [1,5,6](Fig. 2). In con-
tact with a fibrin clot, CPU attenuates fibrinolysis
by removing carboxy-terminal lysines from partially
cleaved fibrin molecules, thereby diminishing its cofac-
tor activity for activation of plasminogen to plasmin
[7–9]. Following its activation, CPU’s activity is unsta-
ble both in vivo and in in vitro experiments (as an
isolated protein), with reported half-lives at 37 °C from
8 to 15 min, hence the U in its name stands for unsta-
ble [10,11]. The inherent and irreversible decay of
CPU’s activity is believed to be of major importance
for its in vivo down-regulation of activity and has been
linked to structural changes of the enzyme [3,12,13].
In vivo, CPU can also be inactivated by proteolytic
degradation, indicating more accessible and flexible
parts of the molecule exist. It was therefore suggested
that the instability of CPU’s activity is due to intrinsic
structural lability of the enzyme, priming its inactiva-
tion [14].
Because of its prominent bridging function between
coagulation and fibrinolysis, the development of CPU
inhibitors as pro-fibrinolytic agents is an attractive
concept [15,16]. But the instability of the enzyme has

prevented crystallization of CPU and the use of struc-
turally based drug design methods. A three-dimen-
sional model of human proCPU based on the structure
of human pancreas procarboxypeptidase B, a closely
related protease exhibiting a higher stability, has been
published recently by Barbosa Pereira et al. [17].
Recently, it was reported independently by two
separate groups that CPU prevents clot lysis from
proceeding into the propagation phase through a
threshold-dependent mechanism [18,19]. The study of
this threshold phenomenon and, more generally, the
study of the effect of CPU on fibrinolysis, are also
severely complicated by its intrinsic instability of
activity.
‘Directed evolution’ approaches allow the random
generation of a large number of mutants followed by
selection for the desired features. Several proteins have
been changed towards more desired properties using this
approach. Some examples include deoxyribonucleo-
side kinases for changed substrate specificities [20,21],
phosphotriesterase for improved catalytic rates [22],
haem peroxidase for exotic environments (for example,
inside a washing machine) [23], or amylase and sub-
tilisin for improved thermostability [24,25].
In this study, we present the generation of CPU
mutants with highly stable activity obtained by
molecular evolution techniques and selection for
decreased thermo-inactivation. To achieve this we used
a directed evolution approach comprising the genera-
tion of random libraries and recombination of advan-

tageous mutations by Fragment-INduced Diversity
(FIND
TM
) technology, as well as site-directed muta-
genesis. A high-throughput screen based on mamma-
lian cells expressing proCPU mutants was developed
to select CPU variants with more thermostable activ-
ity. Seven proCPU mutants were selected and purified.
After activation by thrombin–thrombomodulin, three
showed a remaining activity of more than 80% after
24-h incubation at 22 °C versus 20% for the wild type
(WT), and two of these three showed a more than
25-fold increase in half-life of activity at 37 °C. Using
one of the stable mutants, we have demonstrated the
importance of CPU stability over proCPU concentra-
tion in down-regulating fibrinolysis.
Results
To investigate the role of exposed hydrophobic resi-
dues on the stability of CPU’s activity, 13 point muta-
tions were introduced in proCPU by site-directed
mutagenesis and expressed in 3T3 cells (F135Q, I147S,
F201T, I204Y, I205E, I204Y ⁄ I205E, L214N, F244T,
L281S, L335S, L376Q, T347I). Based on the alignment
of CPU sequence to the structure of carboxypeptidase
Fig. 1. Physiological role of CPU. CPU
attenuates fibrinolysis by removing C-ter-
minal exposed lysines from partially degra-
ded fibrin.
W. Knecht et al. Stable human CPU mutants
FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS 779

B (CPB) [26] (Fig. 2), these mutants were chosen to
replace hydrophobic amino acids of human CPU with
more hydrophilic residues located on the surface of
porcine CPB. In addition, the T347I naturally occur-
ring variation in CPU was reported to double the
half-life (T1 ⁄ 2) of its activity at 37 °C [11] and was
therefore included. We found that the T347I mutant,
when tested in cell culture supernatant, was only 50%
more stable than our WT CPU with threonin at posi-
tion 347 (Table 1). Recently, Barbosa Pereira et al.
[17] proposed, on the basis of their model of human
CPU, that the two consecutive I at positions 204 and
205 are exposed to the surface, and because they are
quite unique to CPU, might be of importance for the
process of CPU’s activity destabilization. When we
changed these two amino acids to their counterpart in
porcine CPB (I204Y ⁄ I205E), the T1 ⁄ 2 of the mutants’
activity was unchanged compared with WT CPU (data
not shown).
In order to create a high number of mutants, ran-
dom mutagenesis was done using either error prone
PCR (EPP) or creating a library of mutants with the
Genemorph PCR mutagenesis kit (GMK, Stratagene,
La Jolla, CA, USA). Sequencing of the full open read-
ing frame (ORF) of randomly picked clones from these
two approaches revealed a base mutation frequency of
0.41 ± 0.22% and 0.55 ± 0.23% per clone in 19
clones from the EPP library and in 17 clones from the
Fig. 2. Multiple alignment of human preproCPU, human preproCPB and porcine proCPB. The amino acid sequences of human preproCPU
(accession number AAP35582.1), human preproCPB (accession number P15086) and porcine proCPB (accession number 1NSA) were

aligned using
CLUSTAL W [40]. The pre- and the propeptide in preproCPU are shaded in black and grey, respectively. Amino acid exchanges
found in mutants with increased thermostability of CPU’s activity are marked in yellow. ‘*’ means that the residues that column are identical
in all sequences in the alignment. ‘:’ means that conserved substitutions have been observed ‘.’ means that semiconserved substitutions
are observed.
Stable human CPU mutants W. Knecht et al.
780 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS
GMK library, respectively. On the amino acid level
this corresponded to an average of 3.7 or 5.2
exchanges per enzyme for the error-prone PCR or the
Genemorph kit, respectively.
In total, 24 600 clones, 14 600 from the EPP library
and 10 000 from the GMK library were screened for
improved thermostability of CPU activity in the super-
natant of mammalian cells in an HTS format. The best
clones selected were more thoroughly analyzed using
an HPLC-based activity assay for CPU. The most sta-
ble clone, GMK1, is five times more stable than WT
CPU. From both libraries, about 1 in every 5000
clones exhibited a more than doubled T1⁄ 2 of activity
compared with WT. The best clones selected from the
random mutagenesis approach, as well as the site-direc-
ted mutagenesis, are summarized in Table 1 and consti-
tute the basis for the first round of FIND
TM
treatment. It was also noted that all clones displayed
fewer mutations than the average number of mutations
present in randomly selected clones from both
libraries.
To explore further combinations of activity stabil-

izing mutations identified in the first screening step
(Table 1) FIND
TM
was used. For the first round of
FIND
TM
approach, the following clones from Table 1
were used in two different combinations: in F1.1:
EPP1, EPP2, GMK1, GMK2 and, in F1.2: all clones
in Table 1 except WT. FIND
TM
libraries were
expressed and 5000 clones of each library screened for
improved thermostability. Table 2 summarizes clones
derived from this step. As shown in Table 2, six clones
with improved T1 ⁄ 2 of their activity compared to the
parental clones could be found in the F1.1 treatment,
while only two clones were found in the F1.2 treat-
ment with improved or equal properties, despite the
higher number of clones put into this library. It should
also be mentioned here that the FIND
TM
treatment
not only recombined existing mutations, but also intro-
duces new mutations as observed in six out of the
eight selected clones (Table 2).
To ascertain the combination of mutations that are
very close in sequential space, the GMK2 clone
(Table 1) was modified by site-directed mutagenesis to
create the mutants YQ and YP, and the T1⁄ 2 of their

activity was determined (Table 2). These combinations
increased the stability of CPU’s activity, especially the
YQ mutant.
Following the first round of FIND
TM
treatment,
50% of the mutants with improved thermal stability
of their activity appeared to bear mutations in the
region encompassing residues 327–357. New mutants
were made by site-directed mutagenesis, trying to
combine the mutations leading to the strongest
decrease in thermo-inactivation by site-directed muta-
genesis. The stability of their activity was evaluated
either after expression in 3T3 cells or in insect cells
using the Baculovirus expression vector system
(BEVS) (Table 3) as an alternative expression system.
The S327P mutation was introduced because P is the
corresponding amino acid to S327 in porcine CPB
(Fig. 2).
A second round of FIND
TM
treatment (F2) then
included the clones: GMK2 + T347I, F1.1.C +
R315H, F1.1.F + S327P, F1.1.A and YQ (see Tables 2
Table 1. Half-life (T1 ⁄ 2) of different CPU mutants’ activity at 37 °C
created by site-directed or random mutagenesis. WT and mutant
CPU were expressed in 3T3 cells and their stability was accessed
in the cell culture supernatant. The remaining enzymatic activity
after incubation of CPU or its mutants at 37 °C was determined
using a HPLC assay.

Clone
Amino acid
changes
in CPU
T1 ⁄ 2at
37 °C
(min)
Method of
generation
EPP1 K166N, H357Q 31 Error prone PCR
EPP2 I251T, H357P 31 Error prone PCR
EPP3 I180F
a
, H357Q 55 Error prone PCR
GMK1 H315R, S327C 60 Genemorph
GMK2 H355Y 47 Genemorph
A L376Q 16 Site-directed
B T347I 18 Site-directed
WT – 12
a
This mutation was not present in all PCR products derived from
this clone.
Table 2. Half-life (T1 ⁄ 2) of different CPU mutants’ activity at 37 °C
derived from the first round of FIND
TM
treatment and site-directed
mutagenesis. WT and mutant CPU were expressed in 3T3 cells
and their stability was accessed in the cell culture supernatant. The
remaining enzymatic activity after incubation of CPU or its mutants
at 37 °C was determined using a HPLC assay. New mutations, not

present in the parental clones are underlined.
Clone Amino acid changes in CPU
T1 ⁄ 2at
37 °C(h)
F1.1.A I251T, H315R, S327C,
N350S, H357Q 2.2
F1.1.B K166N, H315R, S327C,
N350S, H357Q 1.5
F1.1.C K166N, H315R, S327C, H357P 4.4
F1.1.D H315R, S327C,
R352K 1.6
F1.1.E H315R, S327C,
N350S, H357Q 2.4
F1.1.F S327C,
S348N, H357Q 2.9
F1.2.A H315R, S327C, H355Y 2.2
F1.2.B
V219A, H315R, S327C 1
YP
a
H355Y, H357P 1.5
YQ
a
H355Y, H357Q 3
WT – 0.2
a
These mutants were generated by site-directed mutagenesis from
GMK2.
W. Knecht et al. Stable human CPU mutants
FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS 781

and 3). Libraries created from these clones by FIND
TM
technology were expressed, screened and characterized
as described in material and methods. A total of about
14 200 clones were screened. Table 4 summarizes clones
derived from this second round of FIND
TM
. The
same mutation combination as in the best clone made
by site-directed mutagenesis (YQ + S327C) was also
generated by this second round of FIND
TM
treatment
and identified by the screening. The subsequently
increased activity stabilization during the different steps
of directed evolution and screening is illustrated in
Fig. 3, displaying the most stable clones found in each
step.
From the mutants created, seven clones (F1.2.A;
F1.1.F, YQ, YQ + S327C, F1.2.A + R315H,
F1.1.F + N348S, F1.1.F + H355Y) were chosen for
expression using the BEVS and purification of the
mutants for analysis as purified protein. WT proCPU
and mutants were expressed in Sf9 insect cells as
C-terminal His-tagged proteins and purified from the
supernatant of a 1-L culture using IMAC. Figure 4
shows as examples the homogenity of the WT
proCPU-CHis and YQ proCPU-CHis preparations
(0-min samples).
The parameters determined for these mutants are

summarized in Table 5. In contrast to the screening
and previous characterization in crude cell superna-
tants, assays were now carried out in a defined buffer
of 50 mm Hepes, pH 7.4. The T1 ⁄ 2 of CPU activity at
37 °C increased from 0.2 h for WT CPU to more than
5 h for the two most stable clones (Table 5). It appears
that the T1 ⁄ 2 of activity measured directly in the
supernatant of the cell cultures deviates from the T1 ⁄ 2
of the purified proteins in a defined buffer system. It is
likely that cell culture medium components influence
the thermo-inactivation of the mutants. This was con-
firmed by putting purified YQ + S327C back into
insect cell culture medium, which prolonged the T1 ⁄ 2
of activity at 37 °C (data not shown).
A second estimation of the thermal stability of
activity of each mutant was measuring activity after
Table 3. Half-lives (T1 ⁄ 2) of different CPU mutants’ activities at
37 °C made from clones in Tables 1 and 2 by site-directed muta-
genesis. WT and mutant CPU were expressed in 3T3 cells or in
insect cells (as indicated) and their stability was accessed in the cell
culture supernatant. The remaining enzymatic activity after incuba-
tion of CPU or its mutants at 37 °C was determined using a HPLC
assay. The T1 ⁄ 2 of the parental clone is shown in brackets for easy
comparison.
Clone Amino acid changes in CPU T1 ⁄ 2at37°C (h)
GMK2 + T347I
a
T347I, H355Y Not done (0.8)
F1.1.C + R315H K166N, S327C, H357P 1.6 (4.4)
F1.1.A + R315H I251T, S327C, N350S,

H357Q
0.7 (2.2)
F1.2.A + R315H
b
S327C, H355Y 4.3 (2.2)
F1.1.F + N348S
b
S327C, H357Q 2.4 (2.9)
F1.1.F + H355Y
b
S327C, S348N, H355Y,
H357Q
4 (2.9)
F1.1.F + S327P S327P, S348N, H357Q 0.3
c
(2.9)
YQ + S348N S348N, H355Y, H357Q 2.4 (3)
YQ + T347I T347I, H355Y, H357Q 3.5 (3)
YQ + S327P S327P, H355Y, H357Q 1.1 (3)
YQ + N350S N350S, H355Y, H357Q 1.4 (3)
YQ + S327C
b,d
S327C, H355Y, H357Q 26 (3)
WT – 0.2
a
Very low expression level did not allow T1 ⁄ 2 determinations for
GMK2 + T347I.
b
These mutants were expressed in insect cells and
have an 8xHis tag as described in Experimental procedures.

c
Activ-
ity was determined using the Hippuricase assay.
d
The same combi-
nation was independently found within the second FIND
TM
treatment (see Table 4).
Fig. 3. Subsequent increase in stability of activity during the directed
evolution process of CPU. T1 ⁄ 2 data at 37 °C for the most stable
clones as determined in the supernatant of 3T3 cells are presented.
More results for the different steps are presented in the correspond-
ing tables: Random mutagenesis (Table 1), first FIND
TM
(Table 2),
second FIND
TM
⁄ site-directed mutagenesis (Tables 3 and 4).
Table 4. Half-life (T1 ⁄ 2) of different CPU mutants’ activity at 37 °C
derived from the 2nd round of FIND
TM
treatment. WT and mutant
CPU were expressed in 3T3 cells and their stability was accessed
in the cell culture supernatant. The remaining enzymatic activity
after incubation of CPU or its mutants at 37 °C was determined
using a HPLC assay. Mutations not found in the parental clones are
underlined.
Clone
Amino acid changes
in CPU

T1 ⁄ 2at
37 °C (h)
F2.A I251T, H355Y, H357Q 2
F2.B
I204T, Y230C, S348N,
H357Q
2.9
F2.C
a
S327C, H355Y, H357Q 6.8
WT – 0.2
a
Identical to YQ + S327C (see Table 3).
Stable human CPU mutants W. Knecht et al.
782 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS
incubation at 22 °C for 24 h (Table 5). Mutants
YQ + S327C, F1.2.A and F1.1.F + H355Y were
again the most stable and only one mutant, F1.1.F +
N348S, lost more than 50% of its activity.
In order to exclude profound effects of the muta-
tions on enzymatic activity and inhibitor binding affin-
ity, the K
m
for hippuryl-l-arginine (Hip-Arg), the
specific activity at 24 mm Hip-Arg and the IC
50
for the
specific inhibitor PCI were determined. As can be seen
in Table 5, the K
m

values of the mutants shift to lower
values, while all mutants except F1.1.F show an
increased specific activity. To detect any changes in the
positioning of the propeptide, or, in other words, to
see if the contact region between the catalytic domain
and the prodomain was changed by the mutations, we
also measured the residual activity without activation
by thrombin–thrombomodulin. A correct positioning
of the propeptide should keep the residual activity on
Fig. 4. Tryptic digest of WT and YQ proCPU-CHis. (A) SDS PAGE of a bovine trypsin digest of WT proCPU-CHis (1.3 lgÆlane
)1
) and YQ pro-
CPU-CHis (2 lgÆlane
)1
). Two proCPU-CHis to bovine trypsin ratios (w ⁄ w) were used: (i) 1 : 20 and (ii) 1 : 100. Digests were run at 26 °Cfor
the times indicated and then separated by SDS ⁄ PAGE and the gel was Coomassie stained. Two major degradation products of WT- and
YQ-proCPU-CHis became visible and are indicated by arrows in the figure. (B) WT and YQ proCPU-CHis were digested by bovine trypsin as
described under (A) (i) for the times indicated. Fifteen micrograms per lane were separated by SDS ⁄ PAGE and transferred to a polyvinylid-
ene difluoride membrane for N-terminal sequencing (Amidoblack staining). The bands indicated by numbers were identified as starting at the
N-terminus with (i) a mixture of A115 and F23, (ii) a mixture of A115 and F23, (iii) Y353 and (iv) A115.
Table 5. Kinetic and stability parameters for purified WT and mutant CPUs. The T1 ⁄ 2 of activity at 37 °C in cell culture medium is shown in
brackets for easy comparison. Specific activity was determined at 24 m
M Hip-Arg and the IC
50
of PCI at 4 mM Hip-Arg. The specific activity
for 24 m
M Hip-Arg without activation by thrombin–thrombomodulin is given in brackets.
H315 S327 S348 H355 H357
T1 ⁄ 2
at 37

°C(h)
Activity left after
24 h at 22 °Cin
% (mean ± SD)
K
m
(mM)
Specific
activity
(UÆmg
)1
)
IC
50
PTCI
(l
M)
WT 0.2 (0.2) 20 ± 11 2.2 53 (1.9) 0.2
F1.2.A R C Y 5.2 (2.2) 89 ± 8.9 3.7 98 (2.4) 0.04
F1.1.F C N Q 2.2 (2.9) 56 ± 6.3 0.7 41 (1.8) 0.13
YQ Y Q 1.5 (3) 78 ± 7.9 0.9 88 (1.6) 0.16
YQ + S327C C Y Q 5.5 (26; 6.8) 81 ± 7.8 1.1 121 (2.3) 0.16
F1.2.A + R315H C Y 1.3 (4.3) 63 ± 2.2 1.5 150 (2.5) 0.06
F1.1.F + N348S C Q 1 (2.4) 45 ± 3.3 0.6 64 (3.7) 0.12
F1.1.F + H355Y C N Y Q 4.9 (4) 86 ± 6.6 1 89 (4.4) 0.18
W. Knecht et al. Stable human CPU mutants
FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS 783
the same level as for WT-proCPU until it is cleaved
away and proCPU activated to CPU. The activities
ranged from 1.6 to 4.4 UÆmg

)1
, with 1.9 UÆmg
)1
for
WT-proCPU-CHis, or, as a percentage of the specific
activity after activation, from 1.7 to 5.8% with 3.6%
for WT-proCPU-CHis. At 4 mm Hip-Arg as substrate
in the assay, inhibition of all mutants is achieved at
somewhat lower concentrations of potato tuber carb-
oxypeptidase inhibitor (PTCI). The YQ mutant, cho-
sen because it had the most stable activity of the
purified mutants having only two mutations, was used
for further extensive characterization.
To determine if the increased thermal stability of
CPU activity is connected to an increased thermosta-
bility of the protein itself, we monitored the thermal
unfolding of WT and YQ proCPU-CHis. Compared
with the WT, the midpoint temperature (T
m
) of the
protein-unfolding transition has increased for YQ
proCPU-CHis by 10.4 °C (Fig. 5a). Because in YQ
proCPU-CHis, H355 and H357 are replaced by nonio-
nizable amino acids, we monitored thermal unfolding
also at different pH values (Fig. 5b). Approaching low
pH values, when histidines become fully protonated, a
pronounced drop of T
m
was seen with WT proCPU-
CHis, while only a marginal one was recorded with

YQ proCPU-CHis. The drop in T
m
from pH 7.4 to
pH 4.5 was 12.8 °C for WT proCPU-CHis but only
2.3 °C for YQ proCPU-CHis. This indicates a role of
H355 and ⁄ or H357 in the thermal stability of proCPU.
Furthermore, we digested WT proCPU-CHis and YQ
proCPU-CHis with bovine trypsin (Fig. 4), which
resulted in the case of WT proCPU-CHis in one
prominent degradation product of approximately
25 kDa and a weak, probably intermediate band at
about 38 kDa (arrows in Fig. 4A), while for YQ
proCPU-His, a strong band at 38 kDa became visible
but none was visible at about 25 kDa. Subsequently,
N-terminal sequencing of these bands identified a clea-
vage site between R352 and Y353 in WT proCPU-
CHis, but not in the YQ mutant. Consequently, the
two mutations of YQ make the mutant less sensitive
to tryptic digestion close to the positioning of its two
mutations.
Next, we compared the affinity of the enzyme for
synthetic and physiological substrates, and determined
the K
m
constants of native CPU from plasma, recom-
binant WT CPU and YQ CPU for Hip-Arg and bra-
dykinin using an arginine kinase-based kinetic assay
[27]. Data are presented in Table 6. No differences
were seen in the K
m

values of the three CPUs for bra-
dykinin and Hip-Arg when the kinetic assay was used,
proving that the mutations in the YQ proCPU did not
alter the affinity of the carboxypeptidase for synthetic
and physiological substrates. However, when the K
m
for Hip-Arg was measured using HPLC (Table 5), YQ
shows K
m
value similar to the kinetic assay, while WT
CPU does not.
Fig. 5. Thermal unfolding of WT proCPU-CHis and YQ proCPU-
CHis. The thermal unfolding of WT and YQ proCPU-CHis was mon-
itored using the fluorescent dye Sypro orange. The unfolding pro-
cess results in increase in fluorescence, which was monitored. (A)
shows the means of three independent unfolding curves in 50 m
M
Hepes pH 7.4 and the solid line present the best fit of equation 1
to all data. (B) shows the T
m
of thermal unfolding curves at differ-
ent pH values (best fit of equation 1 to all data ± SEM of the fit).
d, YQ proCPU-CHis; s, WT proCPU-CHis. Buffers used were
50 m
M sodium acetate, pH 4.5, 50 mM Mes pH 5.6–6.5, 50 mM
Hepes, pH 7.4.
Table 6. Comparison of K
m
constants of native, WT and YQ CPU
for Hip-Arg and bradykinin using an continuous enzyme assay.

K
m
values are expressed in lMÆL
)1
and are the mean ± SEM of a
duplicate measurement.
Native CPU WT YQ
Bradykinin 39 ± 2 44 ± 6 35 ± 5
Hip-Arg 840 ± 21 825 ± 44 774 ± 39
Stable human CPU mutants W. Knecht et al.
784 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS
The hypothesis that CPU down-regulates fibrinolyis
by a threshold dependent mechanism was recently pub-
lished [18,19]. As long as the CPU activity remains
above this threshold (reported to be 8 UÆL
)1
), fibrinoly-
sis does not accelerate but stays in its initial phase [19].
The study of this threshold phenomenon is severely
complicated by the intrinsic instability of CPU’s activ-
ity. YQ proCPU-CHis was consequently tested for its
antifibrinolytic potential in an in vitro clot lysis assay
and used for confirmation of the threshold hypothesis.
We reconstituted proCPU-depleted plasma with
increasing amounts of the activated stable YQ mutant
or with WT CPU (CPU activities ranging from 0 to
237 UÆL
)1
) and used these in clot lysis experiments, as
described previously [19,28]. Recovery of the added

CPU was in the range of 96–103%, as measured with a
kinetic plasma assay [27]. The final t-PA concentration
used was 40 ngÆmL
)1
. The stable YQ mutant was able
to prolong the in vitro clot lysis time (CLT) in a way that
can be theoretically expected based on its stability.
The decay of CPU can be expressed using the fol-
lowing simplified equation
N ¼ N
0
· e
–k ⁄ t
where k ¼ ln(2) ⁄ T, T ¼ half life of CPU.
Rearrangement of this formula gives the equation:
t ¼ [T log(2)
)1
] · [log(No ⁄ N)], where t is the time
above the threshold, N
0
the initial CPU activity and N
the threshold activity value.
This equation indicates that the time above the
threshold is linearly related with the CPU half life and
only logarithmically with the initial CPU activity (gen-
erated from proCPU by first order kinetics). The hypo-
thesis that this time above the threshold determines the
CLT is strongly confirmed and illustrated in Figs 6
and 7.
Figure 6 shows representative clot lysis profiles at

different YQ CPU concentrations. Increasing the
enzyme activity below the ‘threshold value’ did not
show a significant increase in CLT. However, each
doubling of the CPU activity in excess of the ‘thresh-
old value’ increased CLT with one CPU mutant half
life. Plotting log (CPU activity added) versus CLT
clearly confirms the CPU threshold hypothesis. The
estimated threshold value in our experiments was
12 UÆL
)1
which corresponds very well with the
8UÆL
)1
described by Leurs et al. [19].
Figure 7 illustrates the linear relationship between
CPU stability and CLT. Adding 40 UÆL
)1
WT CPU to
proCPU-depleted plasma increases CLT by 22 min.
However, the addition of 40 UÆL
)1
YQ CPU (with a
7.5- fold increased stability) increases CLT by
153 min, which corresponds very well with the increase
one theoretically can expect (i.e. 7.5 · 22 min). When
the selective CPU inhibitor PTCI (20 lgÆmL
)1
) was
added from the start, no significant prolongation of
CLT was seen by adding YQ or WT CPU.

1 YQ t
1/2
Fig. 6. Threshold hypothesis confirmation. The graph shows representative clot lysis profiles of proCPU depleted plasma reconstituted with
increasing concentrations of activated YQ mutant (concentrations ranging from 0 UÆL
)1
to 237 UÆL
)1
). The threshold value is estimated by
plotting log of the CPU activity added versus the clot lysis time (inset). Each doubling of the enzyme activity above the threshold value
increases clot lysis time with one CPU mutant half-life.
W. Knecht et al. Stable human CPU mutants
FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS 785
Discussion
Due to its physiological role and the need for a very
tight regulation in the blood coagulation cascade, it is
likely that CPU has been selected for intrinsic instabil-
ity, which ensures rapid inactivation of its activity at
the site of action. The irreversible decay in activity has
been shown to be accompanied by structural changes
of CPU [12,13], it is therefore very likely that the loss
of activity is caused by structural changes of the
enzyme triggered upon activation. This instability is a
serious challenge when dealing with overexpression
and purification of the protein. The mechanism behind
CPU’s activity inactivation is still not fully understood,
but several aspects contributing to CPU’s instability
are illuminated by our work.
CPB is a close homologue to CPU, but has a signifi-
cantly higher stability. Aligning the CPU sequence
onto the CPB structure [26] reveals the presence of

numerous potentially exposed hydrophobic amino
acids in CPU. Exposed hydrophobic residues lead to
aggregation, and replacing exposed hydrophobic resi-
dues with more polar residues has been reported to
stabilize proteins [29,30]. Of the 12 hydrophobic to
hydrophilic point mutations carried out in CPU, only
one, L376Q (clone A), had a stabilizing effect, in this
case, of about 33%. All the other mutants either did
not change the T1 ⁄ 2 of CPU’s activity more than
± 20%, or, in the case of I147S, did not express at all
(data not shown), suggesting that the instability does
not result from hydrophobically driven aggregation of
the protein. This is further confirmed by the existence
of a natural variant of CPU, where T347 is subsituted
by an I. Although accentuating the hydrophobic
character of the protein surface, the mutation induces
a stabilization of the protein (Table 1 and [11]).
Random evolution of the enzyme has allowed us to
identify mutants of 2.5 to five-fold increased T1 ⁄ 2in
activity (Table 1), with one or two mutations per clone.
The following first round of FIND
TM
treatment pro-
longed T1⁄ 2 from 12 min for the WT to 4.4 h for clone
F1.1.C. Further combination by rational site-directed
deletion or addition of mutations (Table 3) resulted in
more than half of the cases in a decrease of T1 ⁄ 2. A fur-
ther round of FIND
TM
treatment did not improve T1 ⁄ 2

further compared with a combination of mutations pre-
viously found by site-directed mutagenesis, but inde-
pendently produced the same combination of mutations
that were also determined to display the most stable
activity (YQ + S327C ¼ F2.C). An overall view of the
evolution process is presented in Fig. 3.
A number of mutants with modifications in this
region of the polypeptide chain were expressed in
insect cells, purified and characterized (Table 5). The
mutant displaying the most stable activity at 37 °C
had mutations at the positions S327, H355 and H357,
and this is also reflected by the selection of proteins to
be purified, that all have at least two mutations at
these positions. The T1 ⁄ 2 of activity of the purified
mutants determined in a defined buffer system, as used
during purification procedures, differed significantly
from T1 ⁄ 2 determined in mammalian or insect cell cul-
ture supernatant. From a practical point of view, to
allow for high-throughput mutant screening, thermo-
stability had to be measured in cell culture superna-
tants. The corresponding values obtained from purified
proteins show that cell medium itself and ⁄ or unknown
substances secreted by the cells sometimes strongly
Fig. 7. CPU stability versus proCPU concentration in influencing clot lysis time. The graph shows the effect of adding increasing activities of
WT CPU and YQ CPU on the clot lysis time, clearly showing the importance of the CPU stability over proCPU concentration.
Stable human CPU mutants W. Knecht et al.
786 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS
prolonged or decreased the stability of the activity of
the CPU mutants (Table 5). This is most striking for
YQ + S327C, with T1⁄ 2 of 5.5 h in Hepes buffer,

6.8 h in mammalian cell culture medium and 26 h in
insect cell culture medium. For mutants containing the
S327C mutation, conditions, such as pH, determining
how fast oxidation of the cystein might occur, may
play a role.
Because most of the purification and in vitro assays
are carried out at room temperature, we also determined
the activity after 24-h incubation at 22 °C. Of all puri-
fied mutants, three showed a remaining activity of more
than 80% after 24-h incubation at 22 °C versus 20% for
the WT, and two of these three showed a more than 25-
fold increase in T1 ⁄ 2 of activity at 37 °C. The decreased
K
m
and mostly increased specific activity may partially
reflect the improved stability of activity, especially at
low substrate concentrations during K
m
determinations,
resulting in a higher velocity than for WT CPU and
thereby decreasing the observed K
m
in comparison to
WT CPU. This hypothesis is supported by the use of
a newly developed continuous coupled enzyme assay
instead of the discontinuous HPLC assay that demon-
strated similar K
m
values of the native and WT, and the
YQ mutant CPU with a synthetic and physiological

substrate of CPU. There seem to be no major changes
in the positioning of the propeptide, as indicated by
residual activities of the mutants close to WT-proCPU.
IC
50
values for the inhibition by a specific inhibitor
PTCI [16] are maximally five-fold lower than for the
WT. With the exception of stability of activity, the CPU
mutants appear surprisingly similar to the WT in their
enzymatic properties.
Marx et al. [31] also described the generation of
forms of CPU with a highly stable activity, but in con-
trast to the work presented here, this refers to a hybrid
of CPU ending at position 314 (Fig. 2) fused to the
following C-terminal part of human CPB. This chi-
mera had a half-life of 1.5 h at 37 °C. We therefore
show here that a stabilization of CPU’s activity that is
more than that which naturally occurs can already be
achieved with only one or a few mutations in the
region following position 314 in CPU.
Fifty per cent of the residues mutated in the clones
selected from the first round of FIND
TM
are located
in a distinct region encompassing residues 327–357
(Table 2 and Fig. 2), as well as the naturally occurring
and activity stabilizing mutation T347I. The mutants
with the most stable activity are achieved by combina-
tions of few conservative mutations, S327C, H355Y
and H357Q. Can the effects of the mutations reported

here and the reasons for the increased stability of
activity if connected to structural changes be rationally
explained? The three residues correspond to P300,
Y327 and P329 in porcine CPB (numbering according
to Fig. 2). Keeping a strict orientation of the side
chains, replacing P300 with a serine would leave the
H-bond to the OH group of the side-chain nonsatis-
fied, thereby destabilizing the protein. Based on the
CPB structure, H355 lies in close proximity to a cluster
of charged residues: R324, K326, H330 and E360.
Introducing a Q at position 355 is likely to favour the
formation of H-bonds with one or several of these resi-
dues, attenuating the charge repulsions between some
of the basic amino acids. The stabilization induced by
the replacement of H357 by a Y is more difficult to
explain, but the aromatic nature of the side chain is
likely to interact favourably with the hydrophobic clus-
ter made up of I316, F318, A337 and V341. Another
contribution to the low stability of the WT proCPU is
the close spatial proximity of the three His residues at
330, 355 and 357. In the YQ mutant, two histidines
are replaced by nonionizable amino acids. Although
not very pronounced at physiological pH, partial
charges on the His could induce a destabilizing
charge–charge repulsion effect. This hypothesis is sup-
ported by the findings that WT proCPU-CHis is less
stable in thermal unfolding at low pH, when H330,
H355 and H357 would be protonated, while the drop
of stability of YQ proCPU-CHis is a lot less pro-
nounced (Fig. 5b).

These observations suggest that our mutations
improve residue interactions in this region, leading to
an improved structural stability of the protein. Limited
trypsinolysis of WT and YQ proCPU-CHis further
corroborate this scenario, as trypsin cleavage occurs at
R352 in WT CPU, but not in the mutant harbouring
the H355Y ⁄ H357Q mutations (Fig. 4).
Recently, the hypothesis was put forward that CPU
can down-regulate fibrinolysis through a threshold-
dependent mechanism [19]. We used the stable
YQ CPU mutant to test this hypothesis. The antifi-
brinolytic potential of the stable mutant was tested in
an in vitro clot lysis assay. The YQ mutant was able to
prolong in vitro clot lysis time in a way that can be
expected based on the stability of its activity. Thus YQ
is the first described stable CPU form with conserved
antifibrinolytic potential. This threshold hypothesis
[19] could be confirmed by adding activated YQ pro-
CPU-CHis to proCPU depleted plasma and plotting
CLT versus the log of the CPU activity added. The
threshold value in our experiments was 12 UÆL
)1
,
which is in good agreement to the value reported by
Leurs et al. [19] of 8 UÆL
)1
. As long as CPU remains
above this activity value, fibrinolysis does not proceed
into the acceleration phase. The threshold hypothesis
W. Knecht et al. Stable human CPU mutants

FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS 787
provides a good estimation of why the CPU stability is
a more important factor than the proCPU concentra-
tion in prolonging CLT and the linear relationship
between CPU stability and clot lysis time was
clearly demonstrated (Fig. 7). Because increased antifi-
brinolytic activity and a higher risk of thrombosis can
theoretically be caused by a higher proCPU level or,
more importantly, by increased CPU stability (e.g.
related to the 347 Thr ⁄ Ile polymorphism), the study of
the naturally occurring functional polymorphism at
position 347 should be included in clinical settings
evaluating proCPU as a thrombotic risk factor.
In summary, of seven selected and purified mutants,
three showed a remaining activity of more than 80%
after 24 h incubation at 22 °C versus 20% for the WT;
two of these showed a more than 25-fold increase in
half-life activity at 37 °C. The mutants harbour a lim-
ited number of mutations, presumably on the surface
of the molecule, and present sufficiently similar enzy-
matic activity to be comparable to the WT molecule.
In vitro characterization of YQ in comparison with WT
CPU with respect to activation, physiological CPU
substrates like bradykinin and in clot lysis assays
revealed no differences between mutant and WT CPU,
except for a prolongation of clot lysis time proportional
to the increase in T1⁄ 2 of activity of the mutant. The
YQ mutant was also used to demonstrate the import-
ance of CPU stability over proCPU concentration in
down-regulating fibrinolysis. It is therefore very likely

that the mutants presented here constitute a relevant
model system for structural studies of the enzyme.
Experimental procedures
Cloning of human preproCPU cDNA
The cloning of human preproCPU, e.g. pAM245, has been
described by Stro
¨
mqvist et al. [32].
Directed nucleotide substitutions were introduced into
the preproCPU cDNA with the Quikchange XL site-direc-
ted mutagenesis kit (Stratagene, La Jolla, CA, USA) acc-
ording to the manufacturer’s instructions.
Error-prone PCR was performed according to Cadwell
and Joyce [33,34]. The 100 lL reaction mixture contained 1
fmol of preproCPU cDNA, 50 mm KCl, 10 mm Tris ⁄ HCl,
pH 8.3, 7 mm MgCl
2
, 0.01% (w ⁄ v) gelatine, 0.3 lm of each
primer CPU_fwd_XhoI (atactcgagccaccatgaagctttgcagccttgc)
and CPU_rev_NotI (atcatgcggccgcttaaacattcctaatgacatgc
caag), 0.2 mm dATP, 0.2 mm dGTP, 1 mm dTTP, 1 mm
dCTP, 0.5 mm MnCl
2
, and 2.5 U of AmpliTaq DNA
polymerase (Applied Biosystems, Foster City, CA, USA).
The cycling parameters used were: 94 °C for 2 min,
followed by 30 cycles, each consisting of denaturation at
94 °C for 30 s, annealing at 45 °C for 45 s, and elongation
at 72 °C for 1 min, followed by 72 °C for 7 min.
A second type of random mutagenesis was performed

using the Genemorph PCR mutagenesis kit (Stratagene)
according to the manufacturer’s instructions.
Random recombination of mutated preproCPU
cDNAs
Random recombination of mutated preproCPU cDNAs
was performed using in vitro molecular evolution of protein
function procedure (now known as Fragment-INduced
Diversity (FIND
TM
) technology) according to the methods
disclosed in UK Patent Publication No. GB 2370 038 A
(UK Patent Office, London, UK).
Generation of stable mouse cell lines expressing
proCPU and mutant proCPUs
A retroviral gene delivery and expression system was used to
express proCPU and mutant proCPUs. The WT and mutant
preproCPU cDNA pooled from directed or random muta-
genesis or FIND
TM
treatment were ligated into the multiple
cloning site of a retroviral vector [pFB-neo (Stratagene)].
DNA of the retroviral vectors was transformed into
XL1-Blue electroporation-competent cells (Stratagene)
according to the manufacturer’s instructions. The resulting
colonies were cultured (3 h, 37 °C, 220 r.p.m.) for subse-
quent plasmid purification.
3T3 cells (ATCC, Boras, Sweden) and a MMLV-based
packaging cell line suitable for use with the retroviral vector
[35] were cultured (37 °C, 5% CO
2

) in D5% [Dulbecco’s
modified Eagle’s medium (Sigma, Stockholm, Sweden) sup-
plemented with 5% fetal bovine serum (HyClone, Logan,
UH, USA), heat inactivated at 63 °C for 30 min, and 1%
nonessential amino acids (Invitrogen, Paisley, UK)].
Stable cell lines were generated as described by Krebs
et al. [36]. Briefly, 2.5 lg of the retroviral DNA was transi-
ently transfected into the packaging cell line (80–90%
confluent, 10 cm
2
culture plate, 2 mL D5%) using Lipofec-
tamine 2000 (Invitrogen) according to the manufacturer’s
instructions. The medium was replaced with D5% 5 h post
transfection and 48 h later the virus containing superna-
tants were collected and passed through 0.45 lm filters.
The supernatants (400 lL), together with polybrene (final
concentration 10 lgÆmL
)1
, Sigma), were added to the 3T3
cell line (80–90% confluent, 10 cm
2
culture plate, 2 mL
D5%). The medium was replaced 16 h post infection with
D5% + G418 (D5% supplemented with 0.8 mgÆmL
)1
G418 (Invitrogen) in order to select for stable transfectants.
Following 4–5 days of selection, the cells were cultured
individually in 150 lL D5% + G418 in 96-well plates for
19 days without splitting before expressed proCPU was
analyzed for stability.

Stable human CPU mutants W. Knecht et al.
788 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS
Screening for decreased CPU thermoinactivation
Ten microlitres of supernatant from the cultivation plates
was transferred to 384-well microtitre plates using a
Multimek pipetting robot (Beckman Coulter, Fullerton,
CA, USA). Activation of proCPU to CPU was achieved by
addition of 5 lL (24 nm thrombin from human plasma,
Sigma-Aldrich and 48 nm thrombomodulin from rabbit
lung (American Diagnostica, Stanford, CT, USA) in 20 mm
Hepes pH 7.4 containing 5 mm CaCl
2
and incubated at
room temperature for 10 min. The activation was stopped
by addition of 5 lL20lm phenylalanyl-prolyl-arginyl-
chloromethyl ketone (Calbiochem, Darmstadt, Germany) in
20 mm Hepes pH 7.4 containing 5 mm CaCl
2
.
The thermal stability was assessed by incubating activa-
ted CPU at 37 °C for 20–240 min and determining the
activity of CPU before and after the incubation period
using a colorimetric assay allowing high throughput (hippu-
ricase assay, Professor D. Hendriks, University of Antwer-
pen, Belgium) [37].
Clones that showed significantly higher activity
after the selected time-period than the WT or the best
parental clones used in generation of the libraries
screened were picked and transferred to new 384-well
microtitre plates.

The stability of activity of the re-grown clones was deter-
mined again by incubating the activated CPU at 37 °C and
using the same protocol as used in the primary screening
described above.
Determination of the half-life (T1/2) of CPU’s
activity
The T1 ⁄ 2 of the activity of the best mutants secreted in
the supernatant of the cells in both screens was then
determined as follows: Activated (mutant-) CPU (activa-
tion as described above) was incubated at a constant
temperature and samples were taken after different time-
points. The activity for each sample, y, was plotted
against corresponding time, x, and the T1 ⁄ 2 was then
determined by first fitting equation 503 in Excel Fit
(y ¼ C + A*exp(– B*x)) to the data and then calculating
T1 ⁄ 2 from B (T1 ⁄ 2 ¼ ln2 ⁄ B). Activity was determined
using the HPLC-based assay with hippuryl-l-arginine
(Hip-Arg) as the substrate, as described by Schatteman
et al. [38].
The stability of activity of purified WT and mutant
CPUs were determined with the same protocol, but under
defined buffer conditions of 50 mm Hepes, pH 7.4.
Determination of the ORF of mutated preproCPU
stably expressed in 3T3 cells
After selection of more stable mutant proCPUs (see above),
RNA was purified from selected stable 3T3 cell lines using
Trizol (Invitrogen) according to the manufacturer’s instruc-
tions. Reverse transcription-PCR using preproCPU-specific
primers (CPU_fwd_XhoI and CPU_rev_NotI) were per-
formed with the Titan RT-PCR kit (Roche, Basel, Switzer-

land) according to the manufacturer’s instructions. The PCR
products were subcloned into pGEM-T for sequencing.
Expression of WT proCPU in insect cells
To express WT proCPU, the ORF of preproCPU was
amplified in a PCR reaction using pAM245 as the template
and the following primers: forward: CPU-for1 (tgctctagagcg
gccgcgggatgaagctttgcagccttgcagtccttgtacc); reverse: C-HIS1-
rev (atgatgatgcttatcgtcatcgtccccgggctcgagaacattcctaatga cat
gccaagc) and C-HIS2rev (cggggtaccttattaagatccactatgatga
tgatgatgatgatgatgct tatcgtcatcgtcc).
The resulting PCR fragment was digested with
NotI ⁄ KpnI and ligated into the NotI ⁄ KpnI sites of pFAST-
Bac1 (Invitrogen). The primers C-HIS1rev and C-HIS2rev
introduced the coding sequence for an octa-His tag at the
C-terminus of proCPU (amino acid sequence of the tag:
LEPGDDDDKHHHHHHHHSGS). The resulting plasmid
was named pAM1079. Recombinant Baculovirus for
expression of recombinant proCPU with C-terminal octa-
His tag (proCPU-CHis) was generated starting from
pAM1079 with the Bac-to-Bac
Ò
Baculovirus Expression
System (Invitrogen), according to the manufacturer’s
instructions.
Expression of mutant proCPUs in insect cells
The ORF of selected mutant preproCPUs were amplified
by PCR using the following primers: forward: GateCPUfor
(ggggacaagtttgtacaaaaaagcaggcttcaccatgaagctttgcagcc ttgca
gtccttgtacc); Reverse: C-HIS1rev and C-HIS2rev and Gate-
HISrev (ggggaccactttgtacaagaaagctgggtcctaagatccactatgat

gatgatgatgatgatgatg).
The resulting PCR fragments were subcloned into the
entry vector pDONR201 (Invitrogen) using the Gateway
TM
Technology with help of a BP reaction (Invitrogen) accord-
ing to the manufacturer’s instructions.
Additional site-directed mutagenesis on the inserts within
these plasmids, if desired, was performed as described
above.
Recombinant Baculovirus for expression of recombinant
mutant proCPU with C-terminal octa-His tag was gener-
ated with the BaculoDirect
TM
Baculovirus Expression
System (Invitrogen) according to the manufacturer’s
instructions.
Alternatively, recombinant Baculovirus for expression
of recombinant mutant proCPU was generated with the
Bac-to-Bac
Ò
Baculovirus Expression System (Invitrogen)
according to the manufacturer’s instructions. For this,
pDEST8 (Invitrogen) was used as the destination vector
and the ORF of the mutant was transferred into pDEST8
W. Knecht et al. Stable human CPU mutants
FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS 789
with the help of a LR reaction (Invitrogen) according to
the manufacturer’s instructions.
Purification of WT and mutant proCPUs
expressed in insect cells

SF9 insect cells (Invitrogen) were kept in shaker culture
(27 °C, 105 r.p.m.) in Sf-900II SFM medium (Invitrogen)
and were infected at a multiplicity of infection (MOI) > 1.
The supernatant was harvested after 3–5 days by centrifuga-
tion for 45 min at 6000 g. The supernatant was subsequently
concentrated approximately four times using vivaflow
50 units with a MWCO 10.000 (Vivascience, Hannover,
Germany). The concentrated supernatant was dialyzed over-
night against 50 mm NaH
2
PO
4
, 300 m m NaCl pH 7 (buffer
A). The dialyzed supernatant was loaded on a Talon
TM
Superflow
TM
(Clontech, Mountain View, CA, USA) column.
The column was first washed with five column-volumes of
buffer, then with a gradient up to 45 mm imidazole in buffer
A (five column-volumes), followed by five column–volumes
of 45 mm imidazole in buffer A. Elution of proCPU-CHis
was carried out by a linear gradient (two column-volumes)
from 45 to 125 mm imidazole in buffer A.
ProCPU-CHis containing fractions were pooled and buf-
fer exchanged into 20 mm Hepes, 150 mm NaCl, pH 7.4,
using PD10 columns (Amersham Biosciences, Uppsala,
Sweden) according to the manufacturer’s instructions.
Characterization of purified WT and mutant
proCPUs

SDS ⁄ PAGE was carried out using 4–12% Bis-Tris Gels
(NuPAGE
TM
, Invitrogen) according to the manufacturer’s
instructions. The concentration of proCPU (mutants) in cell
culture supernatants or purified samples was determined
using a proCPU ELISA as described by Stro
¨
mqvist et al.
[39]. T1 ⁄ 2 and activity were determined with the HPLC-
based assay described above.
IC
50
values for the CPU inhibitor PCI (potato tuber
carboxypeptidase inhibitor, Calbiochem) were determined
with activated (mutant) CPUs (activation as described
above) at a substrate concentration of 4 mm Hip-Arg. The
IC
50
values were determined fitting equation 205 in Excel
Fit (y ¼ A + {(B ) A) ⁄ (1 + [(C ⁄ x)
D
]}) to the data, where
y is percentage inhibition and x is the molar concentration
of the inhibitor, A is locked to 0%, B to 100%, C is the
IC
50
value and D is the slope of the curve. The IC
50
value

is the concentration x, where y ¼ 50%.
K
m
was measured using the HPLC assay with activated
(mutant) CPUs (activation as described above) using the
substrate Hip-Arg and the substrate concentration was
plotted against rate (product) formation. The Michaelis-
Menten equation was then fitted to the data using non-
linear fitting in GraFit 4.0. The specific activity of the
purified proteins was determined with 24 mm Hip-Arg. K
m
constants of WT, YQ and native CPU for Hip-Arg and
bradykinin were also determined using a coupled enzyme
assay for CPU activity [27]. One unit of enzyme activity
was defined as the amount of enzyme required to hydro-
lyze 1 lmol of Hip-Arg per minute at 37 °C under the
conditions described.
Further characterization of a selected mutant (YQ)
K
m
constants of YQ CPU-CHis for Hip-Arg and bradyki-
nin were also determined using a coupled enzyme assay for
CPU activity [27] and compared with WT CPU-CHis and
native CPU (purified according to the protocol described
by Schatteman et al. [6]).
Thermal unfolding of WT and YQ proCPU-CHis was
monitored using the fluorescent dye Sypro orange (Molecu-
lar probes) at 10Æ concentration and 15 lgÆmL
)1
protein in a

Jasco FP-6200 Spectroflurometer equipped with ETC-272
Thermocontroller unit. The sample was heated at a heating
rate of 1 °CÆmin
)1
. The fluorescence intensity was measured
with E
x
⁄ E
m
: 490 ⁄ 530. The midpoint temperature (T
m
) of the
protein-unfolding transition was determined by fitting equa-
tion 1 F(T) ¼ min + (max-min) ⁄ (1 + (T ⁄ T
m
)
n
) to the data
using sigma plot 8. F(T) is the fluorescence intensity at
temperature T.
N-Terminal amino acid sequencing
Samples were blotted on polyvinylidene difluoride mem-
brane (Immobilon P, Millipore, Billerica, MA, USA) and
stained with Amidoblack. N-Terminal sequencing was per-
formed on an Applied Biosystems Procise 494 sequencer. A
minimum of eight residues from the N-terminus were ana-
lyzed. The sequencer was controlled by procise2.1 and data
analysis was performed with sequencepro 2.1 software.
Evaluation of the antifibrinolytic potential:
clot-lysis experiments

Polyclonal antiproCPU antibodies from rabbit have been
described previously [39]. Two milligrams of antiproCPU
antibodies were coupled to 0.5 g CNBr-activated Sepharose
(Amersham), as described by the manufacturer. Twenty-five
millilitres of human citrated plasma was incubated with
1 mL of antiproCPU sepharose for 2 h at room tempera-
ture and the gel slurry was then separated from the plasma
by filtration. This was repeated twice more. The depleted
plasma showed no measurable proCPU, as detected by the
reference HPLC-assisted assay.
ProCPU-depleted plasma was reconstituted with increas-
ing amounts of activated WT and YQ proCPU (CPU activ-
ities ranging from 0 to 237 U L
)1
) and used in clot-lysis
experiments, as described by Leurs et al. [19] with some
Stable human CPU mutants W. Knecht et al.
790 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS
minor modifications [28]. Briefly, in each well of a 96-well
microtitre plate, 70 lL proCPU depleted plasma was mixed
with 20 lL t-PA in 20 mm Hepes, 0.1% Tween 20 pH 7.4
(final t-PA concentration 40 ngÆmL
)1
). Forty microlitres
0.9% NaCl and 10 lL activated WT or YQ in 20 mm He-
pes, 5 mm CaCl
2
pH 7.4 were added, and subsequently
clotting was initiated by adding 20 lL 100 mm CaCl
2

and
20 mm Hepes, pH 7.4. Clot-lysis experiments were also per-
formed in the presence of PTCI (final concentration
20 lgÆmL
)1
) added from the start.
Acknowledgements
The authors would like to thank Johanna Deinum and
Petter Bjo
¨
rquist for establishing the contact between
AstraZeneca and Alligator Biosciences. We thank
Marie Karlsson, Patrik Ho
¨
jman, Daniel Herrlander,
Alfia Khairullina and Yani Sim for excellent technical
assistance and Fritz Schweikart for performing the
N-terminal amino acid sequencing. We also thank
Magnus Polla and Laurence D. S. Gainey for helpful
discussion of the manuscript.
References
1 Bouma BN & Meijers JC (2003) Thrombin-acti-
vatable fibrinolysis inhibitor (TAFI, plasma procarboxy-
peptidase B, procarboxypeptidase R, procarboxy-
peptidase U). J Thromb Haemost 1, 1566–1574.
2 Bouma BN & Meijers JC (2004) New insights into
factors affecting clot stability: a role for thrombin acti-
vatable fibrinolysis inhibitor (TAFI; plasma procarb-
oxypeptidase B, plasma procarboxypeptidase U,
procarboxypeptidase R). Sem Hematol 41, 13–19.

3 Nesheim M (2003) Thrombin and fibrinolysis. Chest
124, 33S–9S.
4 Schatteman K, Goossens F, Leurs J, Verkerk R,
Scharpe S, Michiels JJ & Hendriks D (2001)
Carboxypeptidase U at the interface between coagula-
tion and fibrinolysis. Clin Appl Thromb ⁄ Hemost 7,
93–101.
5 Eaton DL, Malloy BE, Tsai SP, Henzel W & Drayna D
(1991) Isolation, molecular cloning, and partial charac-
terization of a novel carboxypeptidase B from human
plasma. J Biol Chem 266, 21833–21838.
6 Schatteman KA, Goossens FJ, Scharpe SS & Hendriks
DF (2000) Proteolytic activation of purified human pro-
carboxypeptidase U. Clin Chim Acta 292, 25–40.
7 Bajzar L, Manuel R & Nesheim ME (1995) Purification
and characterization of TAFI, a thrombin-activable
fibrinolysis inhibitor. J Biol Chem 270, 14477–14484.
8 Redlitz A, Tan AK, Eaton DL & Plow EF (1995)
Plasma carboxypeptidases as regulators of the plasmino-
gen system. J Clin Investig 96, 2534–2538.
9 Wang W, Boffa MB, Bajzar L, Walker JB & Nesheim
ME (1998) A study of the mechanism of inhibition of
fibrinolysis by activated thrombin-activable fibrinolysis
inhibitor. J Biol Chem 273, 27176–27181.
10 Wang W, Hendriks DF & Scharpe SS (1994) Carboxy-
peptidase U, a plasma carboxypeptidase with high affi-
nity for plasminogen. J Biol Chem 269, 15937–15944.
11 Schneider M, Boffa M, Stewart R, Rahman M, Kos-
chinsky M & Nesheim M (2002) Two naturally occur-
ring variants of TAFI (Thr-325 and Ile-325) differ

substantially with respect to thermal stability and antifi-
brinolytic activity of the enzyme. J Biol Chem 277,
1021–1030.
12 Marx PF, Hackeng TM, Dawson PE, Griffin JH, Meij-
ers JC & Bouma BN (2000) Inactivation of active
thrombin-activable fibrinolysis inhibitor takes place by a
process that involves conformational instability rather
than proteolytic cleavage. J Biol Chem 275, 12410–
12415.
13 Boffa MB, Bell R, Stevens WK & Nesheim ME (2000)
Roles of thermal instability and proteolytic cleavage in
regulation of activated thrombin-activable fibrinolysis
inhibitor. J Biol Chem 275, 12868–12878.
14 Marx PF, Dawson PE, Bouma BN & Meijers JC (2002)
Plasmin-mediated activation and inactivation of throm-
bin-activatable fibrinolysis inhibitor. Biochemistry 41,
6688–6696.
15 Zirlik A (2004) TAFI: a promising drug target? Thromb
Haemost 91, 420–422.
16 Lazoura E, Campbell W, Yamaguchi Y, Kato K,
Okada N & Okada H (2002) Rational structure-based
design of a novel carboxypeptidase R inhibitor. Chem
Biol 9, 1129–1139.
17 Barbosa Pereira PJ, Segura-Martin S, Oliva B, Ferrer-
Orta C, Aviles FX, Coll M, Gomis-Ruth FX & Vendrell
J (2002) Human procarboxypeptidase B: three-dimen-
sional structure and implications for thrombin-activata-
ble fibrinolysis inhibitor (TAFI). J Mol Biol 321, 537–
547.
18 Walker JB & Bajzar L (2004) The intrinsic threshold of

the fibrinolytic system is modulated by basic carboxypep-
tidases, but the magnitude of the antifibrinolytic effect of
activated thrombin-activable fibrinolysis inhibitor is
masked by its instability. J Biol Chem 279, 27896–27904.
19 Leurs J, Nerme V, Sim Y & Hendriks D (2004) Carbox-
ypeptidase U (TAFIa) prevents lysis from proceeding
into the propagation phase through a threshold-depen-
dent mechanism.[see comment]. J Thromb Haemost 2,
416–423.
20 Knecht W, Munch-Petersen B & Piskur J (2000) Identi-
fication of residues involved in the specificity and
regulation of the highly efficient multisubstrate deoxyri-
bonucleoside kinase from Drosophila melanogaster. J
Mol Biol 301, 827–837.
W. Knecht et al. Stable human CPU mutants
FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS 791
21 Christians FC, Scapozza L, Crameri A, Folkers G &
Stemmer WP (1999) Directed evolution of thymidine
kinase for AZT phosphorylation using DNA family
shuffling. Nature Biotechnol 17, 259–264.
22 Griffiths AD & Tawfik DS (2003) Directed evolution of
an extremely fast phosphotriesterase by in vitro com-
partmentalization. EMBO J 22, 24–35.
23 Cherry JR, Lamsa MH, Schneider P, Vind J, Svendsen
A, Jones A & Pedersen AH (1999) Directed evolution
of a fungal peroxidase. Nat Biotechnol 17, 379–384.
24 Kim YW, Choi JH, Kim JW, Park C, Cha H, Lee SB,
Oh BH, Moon TW & Park KH (2003) Directed evolu-
tion of Thermus maltogenic amylase toward enhanced
thermal resistance. Appl Environ Microbiol 69, 4866–

4874.
25 Zhao H & Arnold FH (1999) Directed evolution con-
verts subtilisin E into a functional equivalent of thermi-
tase. Protein Engineering 12, 47–53.
26 Coll M, Guasch A, Aviles FX & Huber R (1991) Three-
dimensional structure of porcine procarboxypeptidase
B: a structural basis of its inactivity. EMBO J 10, 1–9.
27 Willemse J, Leurs J, Verkerk R & Hendriks D (2005)
Development of a fast kinetic method for the determi-
nation of carboxypeptidase U (TAFIa) using C-terminal
arginine containing peptides as substrate. Anal Biochem
340, 106–112.
28 Leurs J, Wissing BM, Nerme V, Schatteman K, Bjor-
quist P & Hendriks D (2003) Different mechanisms con-
tribute to the biphasic pattern of carboxypeptidase U
(TAFIa) generation during in vitro clot lysis in human
plasma. Thromb Haemost 89, 264–271.
29 Munoz V, Lopez EM, Jager M & Serrano L (1994)
Kinetic characterization of the chemotactic protein from
Escherichia coli, CheY. Kinetic analysis of the inverse
hydrophobic effect. Biochemistry 33, 5858–5866.
30 Dale GE, Broger C, Langen H, D’Arcy A & Stuber D
(1994) Improving protein solubility through rationally
designed amino acid replacements: solubilization of the
trimethoprim-resistant type S1 dihydrofolate reductase.
Protein Engineering 7, 933–939.
31 Marx PF, Havik SR, Marquart JA, Bouma BN &
Meijers JC (2004) Generation and characterization of a
highly stable form of activated thrombin-activable fibri-
nolysis inhibitor. J Biol Chem 279, 6620–6628.

32 Stromqvist M, Hansson L, Andersson JO, Johansson T,
Edlund M, Enoksson M, Goossens F, Scharpe S &
Hendriks D (2004) Properties of recombinant human
plasma procarboxypeptidase U produced in mammalian
and insect cells. Clin Chim Acta 347, 49–59.
33 Cadwell RC & Joyce GF (1992) Randomization of
genes by PCR mutagenesis. PCR Methods Applic 2,
28–33.
34 Cadwell RC & Joyce GF (1994) Mutagenic PCR. PCR
Methods Applic 3, S136–S140.
35 Miller AD (1997) Development and applications of re-
torviral vectors. In Retroviruses (Coffin M, Hughes SH
& Varmus HE, eds), pp. 437–473. Cold Spring Harbor
Laboratory Press, Cold Spring Harbor, New York.
36 Krebs DL, Yang Y, Dang M, Haussmann J & Gold
MR (1999) Rapid and efficient retrovirus-mediated gene
transfer into B cell lines. Methods Cell Sci 21, 57–68.
37 Schatteman KA, Goossens FJ, Leurs J, Kasahara Y,
Scharpe SS & Hendriks DF (2001) Fast homogeneous
assay for plasma procarboxypeptidase U. Clin Chem
Lab Med 39, 806–810.
38 Schatteman KA, Goossens FJ, Scharpe SS, Neels HM
& Hendriks DF (1999) Assay of procarboxypeptidase
U, a novel determinant of the fibrinolytic cascade, in
human plasma. Clin Chem 45, 807–813.
39 Stromqvist M, Schatteman K, Leurs J, Verkerk R,
Andersson JO, Johansson T, Scharpe S & Hendriks D
(2001) Immunological assay for the determination of
procarboxypeptidase U antigen levels in human plasma.
Thromb Haemost 85, 12–17.

40 Thompson JD, Higgins DG & Gibson TJ (1994) CLUS-
TAL W: improving the sensitivity of progressive multi-
ple sequence alignment through sequence weighting,
position-specific gap penalties and weight matrix choice.
Nucleic Acids Res 22, 4673–4680.
Stable human CPU mutants W. Knecht et al.
792 FEBS Journal 273 (2006) 778–792 ª 2006 The Authors Journal compilation ª 2006 FEBS