Article No. jmbi.1999.3447 available online at http://www.idealibrary.com on
J. Mol. Biol. (2000) 296, 651±658
Electrostatic Dependence of the ThrombinThrombomodulin Interaction
Abel Baerga-Ortiz1, Alireza R. Rezaie2 and Elizabeth A. Komives1*
1
Department of Chemistry and
Biochemistry, University of
California, San Diego, La Jolla
CA 92093-0359, USA
2
Department of Biochemistry
and Molecular Biology, St
Louis University School of
Medicine, 1402 South Grand
Blvd., St Louis, MO
63104, USA
The rate constants for the binding interaction between thrombin and a
fully active fragment of its anticoagulant cofactor, thrombomodulin, have
been determined by surface plasmon resonance. At physiological ionic
strength, the ka was 6.7 106 Mÿ1 sÿ1 and the dissociation rate constant
was 0.033 sÿ1. These extremely fast association and dissociation rates
resulted in an overall binding equilibrium constant of 4.9 nM, which is
similar to previously reported values. Changing the ionic strength from
100 mM to 250 mM NaCl caused a tenfold decrease in the association
rate while the dissociation rate did not change signi®cantly. A similar
effect was observed with tetramethylammonium chloride. A DebyeHuÈckel plot of the data had a slope of ÿ6 and an intercept at 0 ionic
strength of 109 Mÿ1 sÿ1. The same slope and intercept were obtained for
data that was collected in the presence of glycerol to slow the association
rates. These results show that the thrombin-TM456 interaction is extremely rapid and nearly completely electrostatically steered. An association
model is presented in which TM456 approaches thrombin along the
direction of the thrombin molecular dipole.
# 2000 Academic Press
*Corresponding author
Keywords: streptavidin; biotin; biosensor; biacore; electrostatic steering
Introduction
Thrombin generation at the site of injury must
be rapid and controlled. Only small amounts of
thrombin are generated and its action must be
halted immediately in order for hemostatic balance
to be maintained. Thrombomodulin (TM) is
attached to the endothelial cell surface. It captures
thrombin from the ¯owing blood and the thrombin-TM complex initiates the anticoagulant pathway that shuts down thrombin production. The
thrombin-TM complex no longer recognizes ®brinogen, but instead has new catalytic activity
towards protein C. The resulting activated protein
C inactivates the essential limiting cofactors
responsible for thrombin generation (Esmon, 1995).
Although TM is a 70 kDa protein, a small region
containing only three epidermal growth factor-like
(EGF) domains (TM456) is a fully active anticoagulant cofactor for thrombin (Tsiang et al., 1992;
White et al., 1995). The three essential EGF-like
Abbreviations used: EGF, epidermal growth factor;
TM, thrombomodulin; SA, streptavidin; SPR, surface
plasmon resonance; H/2H exchange, hydrogen/
deuterium exchange.
E-mail address of the corresponding author:
ekomives@ucsd.edu
0022-2836/00/020651±8 $35.00/0
domains of TM are located in the extracellular
portion of the protein and they must attract and
bind thrombin in this non-equilibrium, ¯owing
system.
Several pieces of evidence indicate that the
thrombin-TM456 interaction is weakened at high
ionic strength. The binding of full-length TM or
TM456 to thrombin that results in inhibition of
®brinogen cleavage decreases as the concentration
of NaCl increases (Vindigni et al., 1997). Similarly,
complexation of thrombin with full-length TM or
TM456 and subsequent protein C activation
decreases with increasing concentration of NaCl
(A. Baerga-Ortiz, unpublished data). The electrostatic surface potential of thrombin reveals a positively charged region known as the anion binding
exosite I, which has been proposed as the TMbinding site. Mutation of residues in this region
results in decreased af®nity for TM (Wu et al.,
1991; Tsiang et al., 1995). The X-ray crystal structure of a nona-decapeptide from the ®fth EGF-like
domain of TM complexed to thrombin shows that
this peptide binds in anion-binding exosite I
(Mathews et al., 1994). Molecular docking studies
show that binding of this same TM peptide to
thrombin changes the electrostatic surface potential
of thrombin, suggesting that the formation of the
complex may be governed, at least partially, by
# 2000 Academic Press
652
electrostatics (Sampoli Benitez et al., unpublished
data).
The in¯uence of electrostatics has been analyzed
for several different protein-protein interactions
(Raman et al., 1992; Schreiber & Fersht, 1996; Radic
et al., 1998; Perrett et al., 1998; Xavier et al., 1998,
1999). Four of the interactions studied were antibody-antigen complexes and two were enzymeinhibitor complexes. Perrett et al. (1997) studied the
interaction of unfolded barnase with GroEL. For all
the interactions that have Kd values in the nanomolar range, the dissociation rates were slow (on the
order of 10ÿ4-10ÿ5 sÿ1), and unaffected by changes
in ionic strength. The association rates as measured
by either ka or kobs, decreased with increasing ionic
strength. The log ka decreased linearly with the
square root of ionic strength in accord with DebyeHuÈckel theory.
The interaction of thrombin with TM differs in
a signi®cant way from the previously studied
interactions for which the functional goal is irreversibility. The thrombin-TM interaction is an
enzyme-cofactor interaction that should be
rapidly reversible so that the cofactor can activate
many enzyme molecules. In the case of the
thrombin-TM interaction, a third player, the irreversible inhibitor antithrombin III, may inhibit
thrombin while bound to TM. The rapid disengagement of TM from the inhibited thrombin is
thus also critical. We present real time kinetic
measurements from surface plasmon resonance
(SPR) experiments carried out at different NaCl
and (CH3)4NCl concentrations that allow quantitative measurement of the electrostatic effects of
the thrombin-TM456 interaction. The results presented here agree with previous equilibrium binding experiments and show that the thrombin-TM
interaction is rapid and highly electrostatically
steered.
Results
Thrombin and thrombin inhibited with PPACK
show identical kinetics of interaction with
biotinylated TM456 coupled to the SA
sensor chip
In order to obtain a homogeneous distribution of
protein molecules on the surface of the SPR sensor
chip, a coupling strategy involving biotinylation of
TM456 was chosen. Fortuitously, TM456 contains
no lysine residues, so biotin with an LC-LC linker
arm was covalently attached only to the N terminus of TM456. Thus, all TM456 molecules bound
to streptavidin (SA) on the sensor chip surface
were linked to the surface in the same manner
and extended far enough from the SA surface to
allow protein-protein interaction to occur. Indeed,
thrombin interaction with TM456 bound to the SA
chip with the shorter biotin-LC-linker was
impaired.
Nanomolar concentrations of thrombin in buffer
containing 150 mM NaCl were ¯owed over a sur-
The Thrombin-Thrombomodulin Interaction
face containing 370 resonance units of immobilized
biotinylated TM456. Identical curves for association
and dissociation were obtained for active thrombin
compared to thrombin that had been inactivated
with D-Phe-Pro-Arg -chloromethylketone (PPACK)
(data not shown). Very similar kinetics were also
observed for human thrombin. For the thrombinTM456 interaction, pseudo ¯owing equilibrium
was attained even at concentrations of thrombin
below the expected Kd, and the ka and kd were
extremely fast compared to most protein-protein
interactions. In order to guard against mass
transport limitation under these rapid binding
conditions, very low amounts of TM456 were
immobilized on the surface, and the highest possible ¯ow rates were used.
One dif®culty in studying the ionic strength
dependence of the thrombin-TM interaction is that
thrombin binds sodium with a binding af®nity in
the millimolar range (Wells & DiCera, 1992).
Sodium binding alters the enzymatic activity of
thrombin as well as the binding af®nity of TM
(Vindigni et al., 1997). Measurement of binding
kinetics at different ratios of NaCl to (CH3)4NCl
keeping the ionic strength constant gave identical
values of ka and kd regardless of sodium concentration (data not shown). From this we concluded
that PPACK-inactivated thrombin did not show a
sodium effect on the interaction with TM456. By
using PPACK-inactivated thrombin for the binding
kinetics measurements, it was possible to study
solely the electrostatic effects of altering the NaCl
concentration without complications from sodiumbinding phenomena.
Effects of NaCl and (CH3)4NCl on the
binding kinetics
Binding kinetic constants for the thrombinTM456 interaction were determined at ®ve different thrombin concentrations and at different
concentrations of NaCl and (CH3)4NCl ranging
from 100 to 250 mM. Examples of the global ®ts
for a set of data from lower ionic strength
(150 mM) and higher ionic strength (250 mM) are
shown in Figure 1. The residuals were low, and
highly reproducible kinetic constants were
obtained from global ®ts of all data sets. Data from
a representative set of data collected at different
concentrations of NaCl are presented in Table 1
and at different concentrations of (CH3)4NCl in
Table 2. Three different experiments were performed on two different Biacore SPR instruments
with different preparations of proteins, and the
results were very reproducible. Within error, the kd
remained constant at approx. 0.03 sÿ1 (Figure 2).
The ka decreased an order of magnitude from
1.8 107 Mÿ1 sÿ1 to 1.8 106 Mÿ1 sÿ1 as the ionic
strength was increased from 100 to 250 mM
(Figure 3).
Whether the ionic strength was increased by
addition of NaCl or (CH3)4NCl, the log of the
association constant decreased linearly with
653
The Thrombin-Thrombomodulin Interaction
Figure 2. Plot of log kd versus the ionic strength from
100 to 250 mM using either (*), NaCl or ( & ),
(CH3)4NCl. Each was the average of at least two experiments of globally ®tted data from ®ve different
PPACK-thrombin concentrations as described in
Materials and Methods. The error bars are the standard
deviation from the mean for the determinations. One
example of the raw data is shown in the inset.
Figure 1. Sensorgrams of PPACK-thrombin in 10 mM
Hepes buffer with (a) 150 mM NaCl or (b) 250 mM
NACl ¯owed at 100 ml/minute over 370 RU of biotinylated TM456 bound to the streptavidin surface. Thrombin concentrations were 0.78 (black), 1.56 (blue), 3.125
(green), 6.5 (orange) and 12.5 nM (red) for the 150 mM
NaCl experiment. Thrombin concentrations were 6.0
(orange), and 12.5 nM (red), 25.0 (cyan) and 50.0 nM
(purple) for the 250 mM NaCl experiment. The data
were ®t globally using the Langmuir 1:1 binding model
in Biaevaluation 3.0. Control experiments in which the
same concentrations of thrombin were ¯owed over a
surface containing only biotin were subtracted from the
sample data.
increasing ionic strength and the data ®t a simple
Debye-HuÈckel model giving a slope of ÿ6
(Figure 4). Because the ka for the thrombin-TM456
interaction was so high and subject to limitations
of measurement by SPR, con®rmatory experiments
were also performed at various glycerol concentrations. Glycerol was expected to increase the viscosity and therefore decrease the ka measured at
each ionic strength. Shown also in Figure 4 are the
data obtained at 20 % (w/v) glycerol. The slope of
all the Debye-Huckel plots was ÿ6. The intercept
at 0 glycerol and ionic strength was 109 Mÿ1 sÿ1
indicating that the association of thrombin with
TM456 is highly electrostatically steered.
Discussion
Protein-protein interactions can be classi®ed
according to their functional consequences. Antibody-antigen interactions, such as the well-studied
HYEL-5-lysozyme interaction, result in removal of
the antigenic protein from the circulation. Enzymeinhibitor interactions such as acetylcholinesterasefasciculin and barnase-barstar result in inhibition
of enzymatic activity. The functional goal of these
types of protein-protein interactions is irreversibility, and therefore the kd for these interactions is
slow. On the other hand, enzyme-cofactor interactions such as the thrombin-TM456 interaction,
should be rapidly reversible so that the cofactor
can activate many enzyme molecules. Thus, it is
not surprising that the kd for the thrombin-TM456
interaction is several orders of magnitude faster,
approximately 0.03 sÿ1, than reported kd values of
10ÿ4-10ÿ5 sÿ1 for functionally inhibitory interactions (Table 3). In order to achieve high af®nity
binding, the ka for an enzyme-cofactor interaction
must be commensurately high; 107 Mÿ1 sÿ1 for the
thrombin-TM456 interaction. In theory, large proteins the size of thrombin and TM would undergo
many unproductive encounters and since few collisions would be productive, the resulting ka would
be slow (Janin, 1997; McCammon, 1998). Electro-
654
The Thrombin-Thrombomodulin Interaction
Table 1. Kinetic constants for PPACK-thrombin binding to 370 response units of TM456
NaCl (mM)
100
125
150
175
200
225
250
ka (Mÿ1 sÿ1)
7
1.5 10
2.2 107
6.7 106
4.9 106
2.2 106
2.2 106
1.8 106
kd (sÿ1)
Kd (nM)
w2
Rmax
0.026
0.025
0.033
0.030
0.029
0.034
0.062
1.8
1.1
4.9
6.1
13.4
15.8
35.3
2.03
0.78
0.47
1.17
0.49
0.19
1.62
42.1
20.2
22.6
20.2
24.9
20.2
17.5
Table 2. Kinetic constants for PPACK-thrombin binding to 370 response units of TM456
(CH3)4NCl (mM)
100
125
150
175
200
225
250
ka (Mÿ1 sÿ1)
7
1.8 10
1.2 107
8.1 106
6.8 106
4.8 106
2.2 106
2.3 106
kd (sÿ1)
Kd (nM)
w2
Rmax
0.021
0.020
0.023
0.022
0.033
0.043
0.044
1.2
1.6
2.8
3.2
6.8
20.0
19.0
2.53
2.61
2.46
1.38
0.89
0.80
1.25
37.7
29.3
24.9
22.5
23.8
26.1
23.4
static steering is one effective way to increase the
probability of productive encounters and achieve
higher ka values and higher interaction af®nities.
SPR provides an excellent way to measure rapid
protein-protein interaction kinetics in a non-equili-
brium ¯owing system. Data can be collected at a
5 Hz sampling rate, and at a ¯ow rate of 100 ml/
minute. Furthermore, the ka and the kd can be accurately measured under the same conditions (Pearce
et al., 1996; Schuck & Minton, 1996; for reviews, see
Raghavan & Bjorkman, 1995; Szabo et al., 1995). A
gauge of the reliability of the kinetic values
obtained can be seen from comparison of the ionic
strength dependence of the Kd obtained by SPR
compared to that measured by binding competition (Figure 5).
Figure 3. Plot of log ka versus the ionic strength from
100 to 250 mM using either (*), NaCl or ( & ),
(CH3)4NCl. Each was the average of at least two experiments of globally ®tted data from ®ve different
PPACK-thrombin concentrations as described in
Materials and Methods. The error bars are the standard
deviation from the mean for the determinations. One
example of the raw data is shown in the inset.
Figure 4. Debye-HuÈckel plot of the log ka versus the
square root of the ionic strength. The ionic strength was
varied from 125 to 250 mM using either (*), NaCl or
( & ), (CH3)4NCl. In a separate experiment, the ionic
strength was varied from 125 to 250 mM using NaCl in
the presence of 20 % glycerol (^). Each experiment was
repeated and the error bars are the standard deviation
from the mean for the two determinations.
655
The Thrombin-Thrombomodulin Interaction
Table 3. Summary of studies on the ionic strength dependence of various protein-protein interactions
Protein
Cyt. c (ox)/mAb2B5
Cyt. c (ox)/mAb5F8
Barnase/Barstar
HE-lysozyme/mAbHY-5
AchE/Fasciculin
BWQ-lysozyme/mAbHY-5
Thrombin/TM456
kd (sÿ1)a
ÿ5
3-12 10
1 10ÿ4
1.5 10ÿ5
2.2 10ÿ4
7 10ÿ5
0.96
3.3 10ÿ2
ka (Mÿ1 sÿ1)
5
6.5 10
1.5 106
1.35 106
1.5 107
2 106
1.8 107
6.7 106
Debye-Huckel
slope
ÿ2
NDb
ÿ2.5
ÿ0.5
ÿ4.8
ÿ0.5
ÿ6
Reference
Raman et al. (1992)
Raman et al. (1992)
Schreiber & Fersht (1993)
Xavier et al. (1998)
Radic et al. (1998)
Xavier et al. (1999)
This report
We apologize if some studies were not included here.
The values reported for kd and ka are those measured at 150 mM ionic strength.
b
ND means not determined.
a
One interesting result from these experiments is
that the thrombin-TM456 interaction kinetics indicate a very high degree of electrostatic steering.
The slopes of all the Debye-HuÈckel plots were
highly reproducible with an average value of
ÿ6.0 ‡/ÿ 0.6, which is the largest ever reported
for a protein-protein association. The Einstein-Smoluchowski equation predicts that the rate of
encounter for two spheres in water at 300 K is
independent of the size of the spheres and
is approx. 6.6 109 Mÿ1 sÿ1. Extrapolation of the
Debye-HuÈckel plot for the thrombin-TM456 interaction gives an association rate constant at zero
ionic strength of 109 Mÿ1 sÿ1, a value that
approaches the Smoluchowski limit. It therefore
appears that the thrombin-TM456 interaction has
approached the limit of electrostatic steering, and
that almost every encounter is a productive one.
In trying to rationalize these results on the basis
of the structure of thrombin, a second interesting
result was attained. The structure of the thrombinTM456 complex is not known, and thrombin has
two highly positively charged patches on the surface, each of which might attract the TM456 molecule. We recently completed hydrogen/deuterium
Figure 5. Comparison of the dependence of log Kd
versus the square-root of the ionic strength for the
thrombin-TM456 interaction determined by (*), inhibition competition assays (Vindigni et al., 1997) and ( & ),
determined in this report by SPR.
(H/2H) exchange experiments that provided information about sites on the thrombin molecule
where TM456 binds. These experiments showed
two regions of thrombin that interact with TM
fragments; the anion-binding exosite I and a loop
near the active site containing the sequence
WRENL (residues 96-100 in the chymotrypsin
numbering system) (Mandell et al., 1998). This
``footprint'' of TM45 on thrombin is shown in
Figure 6(a). The two discontinuous sites of TM
interaction are each near one of the positively
charged patches. Based on these H/2H exchange
experiments, we proposed that TM456 binds in a
manner that connects these two regions (Mandell
et al., 1998). Interestingly, the positive end of the
overall molecular dipole of thrombin extends from
the molecule exactly in the middle of the proposed
binding site that connects the two discontinuous
footprints of TM456 (Figure 6(b)).
This intriguing result begs the hypothesis that
TM456 approaches thrombin along the direction of
the molecular dipole, and that this direction of
approach would position TM456 perfectly for binding. To test this hypothesis, we measured the binding kinetics of TM456 to two mutant thrombins.
The R97A mutation lies on a surface loop outside
the active site and makes contacts to ®brinopeptide
A (He et al., 1997) and is in the WRENL loop that
showed slowed H/2H exchange in the thrombinTM45 complex (Mandell et al., 1998). The K70D
mutation is in anion-binding exosite I and in structures of thrombin, it appears to be participating in
a salt bridge with E80. Calculation of the molecular
dipole of each of the mutant thrombins showed
that for one, the R97A mutant, only the magnitude
of the dipole changed but the direction remained
constant. For the other, the K70D mutant, both the
direction and the magnitude of the dipole changed.
One would predict that if the direction of the
thrombin molecular dipole were important, the
effect of the K70D mutation on binding of TM456
would be more drastic. Furthermore, one would
predict that both the ka and kd would change
equally for the R97A mutant. Both of these predictions were shown to be true (Table 4). Clearly
more work needs to be done to establish the role
of individual residues on TM binding to thrombin
as well as to con®rm the molecular dipole hypoth-
656
The Thrombin-Thrombomodulin Interaction
et al., 1995). Biotinylation was carried out using the EZLink sulfoNHS-LC-LC biotinylation kit from Pierce
Chemicals (Rockford, IL) according to the instructions
provided with the kit. Typically, 90 mg of lyophilized
TM456 prepared by dissolution in 180 ml of H2O was
reacted with 21 ml of 0.018 M NHS-LC-LC-Biotin for two
hours at 4 C. Reaction products were puri®ed on a
Vydac C18 (4.6 mm 250 mm) analytical reverse-phase
HPLC column at a ¯ow rate of 1 ml/minute of 100 %
buffer A (0.1 % tri¯uoroacetic acid) for ten minutes followed by a gradient consisting of 0.1 % tri¯uoroacetic
acid to 50 % (v/v) acetonitrile over 45 minutes. A single
peak was recovered and assayed both for TM456 activity
(White et al., 1995) and for the degree of biotinylation by
the 2-(40 -hydroxyazo-benzene) benzoic acid assay as
described in the biotinylation kit. The yield of puri®ed,
modi®ed TM456 was approximately 35 %. The biotin
label did not have any effect on the cofactor activity of
TM456 in a protein C activation assay.
Coupling of biotin-labelled TM456 to the sensor chip
Figure 6. (a) Structure of thrombin showing the
regions of thrombin for which solvent exclusion
occurred upon binding of TM45. TM45 is a two-EGF
domain fragment of TM that binds somewhat more
weakly to thrombin than TM456, the three EGF-domain
fragment. (b) Electrostatic potential map for thrombin
showing the overall molecular dipole. The thrombin
molecules shown in (a) and (b) are in the same orientation.
esis. Importantly, mutations far from the proposed
TM binding surface need to be studied. Nevertheless, kinetic, structural, and mutational data are so
far consistent with the idea that the thrombinTM456 interaction is highly electrostatically steered
along the direction of the thrombin molecular
dipole.
Materials and Methods
Synthesis of biotin-labelled TM456
TM456 was expressed in the Pichia pastoris yeast
expression system and puri®ed to homogeneity as
described (White et al., 1995). TM456 shows full TM
activity in solution in both thrombin binding assays and
protein C activation assays (Vindigni et al., 1997; White
Coupling of TM456 to the gold surface of the sensor
chip was achieved by ¯owing 1 mg/ml (59 nM) of biotin-labeled TM456 in 10 mM Hepes buffer, 300 mM
NaCl, 2.5 mM CaCl2 (pH 7.4), over a SA sensor chip
(Biacore Inc., Piscataway, NJ). The amount of TM456
immobilized was determined after washing to be 370
units. Coupling was carried out at 300 mM NaCl in
order to collapse the dextran matrix and thus minimize
artifacts that might occur during the experiments at
different ionic strengths. Because the TM456 is highly
glycosylated, this value can not be easily translated into
an amount of protein on the chip. Surfaces were also prepared with only 150 units of immobilized TM456 and
the interaction kinetics measured at these very low
amounts of TM456 were the same as at 370 resonance
units. A control surface was prepared by ¯owing free
biotin (0.003 mg/ml) over a second channel of the SA
sensor chip and data from this ``blank'' channel was subtracted from the sample data.
Preparation of thrombin
Bovine thrombin was prepared as described by Ni
et al. (1990) and puri®ed by chromatography on a
MonoS 10/10 column using a gradient of 100 mM to
500 mM NaCl in NaKPO4 buffer (pH 6.5). Thrombin
eluted from the column at approximately 250 mM NaCl,
and was concentrated to 1.5 mg/ml and stored in small
portions at ÿ70 C. Human thrombin was a generous
gift of J. Fenton, and was puri®ed as described for the
bovine thrombin before use.
During the SPR experiments, we discovered that
thrombin, both active and after inactivation with PPACK
was relatively unstable. Thrombin that had been stored
at ÿ20 C for more than three weeks, or had been dialyzed was unsuitable for SPR analysis due to the high
bulk refractive index of the ¯owing thrombin solution,
perhaps indicating aggregation had occurred. Although
the active thrombin was unsuitable for SPR experiments,
it remained fully active in both ®brinogen cleavage and
protein C activation assays. In order to avoid high bulk
refractive index corrections, all of the thrombin used in
the experiments presented here was freshly prepared.
Active thrombin was stored frozen at ÿ70 C for less
than three weeks before use. This thrombin was thawed,
treated with PPACK, puri®ed by chromatography on a
657
The Thrombin-Thrombomodulin Interaction
Table 4. Summary of studies on the relationahip between thrombin molecular dipole and interaction kinetics
Protein
Thrombin
Thrombin K70D
Thrombin R97A
Difference in dipole angle
(deg.)a
Difference in dipole
magnitude (Debyes)
ka(Mÿ1sÿ1)b
kd(sÿ1)
0
9.8
0.4
0
ÿ52
ÿ32
8.7 106
NDc
3.2 106
0.035
ND
0.069
a
The program Del Phi was used to calculate the overall molecular electrostatic potential grid from which the dipole was calculated in grasp. The dipole is de®ned byx, y, and z coordinates and a magnitude in Debyes. The difference in angle was calculated as
cos y ˆ ab/jajjbj.
b
The values reported for kd and ka are those measured at 150 mM ionic strength.
c
ND means no detectable binding was observed.
Mono S 10/10 column as described above except that
the gradient salt was (CH3)4NCl. The PPACK-inactivated
thrombin was stored at ÿ70 C and used within one
week.
Thrombin mutants were prepared as described (He
et al., 1997). Both the K70D and the R97A mutations
have good speci®c activities. All mutations are denoted
by the chymotrypsin numbering scheme, which is different from the sequential numbering scheme used in
Mandell et al., 1998. In the sequential numbering scheme,
K70 corresponds to K101 and R97 corresponds to R129.
Sensorgrams for thrombin interaction with
biotinylated TM456 coupled to the SA sensor chip
Sensorgrams were collected for thrombin (0.78 nM,
1.56 nM, 3.125 nM, 6.25 nM, 12.5 nM) ¯owed over a sensor chip containing SA-biotin-TM456 or SA-biotin in
10 mM Hepes buffer, 150 mM NaCl, 2.5 mM CaCl2
(pH 7.4). Preliminary experiments revealed a ¯ow rate
dependence on the dissociation rate constant in which
¯ow rates less than 50 ml/minute showed increasing kd
with increasing ¯ow rates. All experiments were therefore carried out at the maximum ¯ow rate of 100 ml/
minute and at a sampling rate of 5 Hz on a Biacore 3000.
Because the dissociation rate was so fast, complete dissociation occurred after three minutes and no surface
regeneration was required.
Rate constants for association (ka) and dissociation (kd)
were obtained by globally ®tting the data from ®ve injections of thrombin using the BIAevaluation software
version 3.0. using the simple Langmuir binding model.
Statistical analysis of the curve ®ts for both dissociation
and association phases of the sensorgrams show low w2
values (0.2-2.6), and low residuals.
Determination of ka and kd at different NaCl and
(CH3)4NCl concentrations
Binding kinetic constants were determined at different
concentrations of NaCl and (CH3)4NCl. The ¯owing buffer was made 100 mM, 125 mM, 150 mM, 175 mM,
200 mM, 225 mM or 250 mM in NaCl or (CH3)4NCl. The
thrombin samples that were injected were prepared in
each different buffer so that the ¯owing buffer and the
thrombin sample buffer were always identical. A total of
®ve different concentrations of thrombin were injected
for every salt concentration. Since the af®nity of thrombin for TM456 decreases with increasing ionic strength,
higher concentrations of thrombin were used in order to
bracket the Kd for the thrombin-TM456 binding at high
salt. For the 100 mM NaCl (or (CH3)4NCl) experiment,
0.39 nM, 0.78 nM, 1.56 nM, 3.125 nM and 6.25 nM of
thrombin were injected. For the 125 mM, 150 mM and
175 mM NaCl (or (CH3)4NCl) experiment, 0.78 nM,
1.56 nM, 3.125 nM, 6.25 nM and 12.5 nM of thrombin
were injected. For the 200 mM NaCl (or (CH3)4NCl)
experiment, 3.125 nM, 6.25 nM, 12.5 nM, 25 nM and
50 nM of thrombin were injected. Finally, for the
250 mM NaCl (or (CH3)4NCl) experiment, 6.25 nM,
12.5 nM, 25 nM, 50 nM and 100 nM of thrombin were
injected. The concentration of the thrombin stock
solution was high enough that dilutions were at least
1:100 and the concentration of (CH3)4NCl in the thrombin stock buffer added a negligible amount of ions to the
®nal buffer.
Determination of the electrostatic dependence of the
ka and kd at 20% glycerol
Kinetic parameters were measured at NaCl concentrations of 125 mM, 150 mM, 175 mM and 250 mM at
20 % glycerol. The wider linear range of the Biacore 3000
accommodates the larger bulk refractive index of the glycerol solutions. Since the addition of glycerol results in a
lower af®nity of thrombin for TM, higher thrombin concentrations were injected in order to achieve saturation
of the surface. For the 125 mM, 150 mM, and 175 mM
NaCl experiments, 3.125 nM, 6.25 nM, 12.5 nM, 25 nM
and 50 nM thrombin was injected. For the 250 mM NaCl
experiment, 12.5 nM, 25 nM, 50 nM, 100 nM and
200 nM thrombin was injected. The ka and kd values
were obtained as already described.
Calculation of the molecular dipole of thrombin
The surface potential for thrombin was calculated
using the Poisson-Bolzmann equation implemented in
DelPhi. A grid of dimensions (65 65 65) and a
Ê was generated. The structure of
spacing of 1.0 A
thrombin used in the calculations was as described by
Bode et al. (1992). Solvent dielectric constant was set
at 80.0 for the solvent and 2.0 for the solute. The
ionic strength was set at 150 mM. The molecular
dipole was calculated from the surface potential using
GRASP (Nichols et al., 1991).
Acknowledgments
We thank Dr Susan Taylor for the use of her Biacore
instrument in the early part of this work, and Ken Dickerson and Shirley Demer for extensive help in setting-up
658
The Thrombin-Thrombomodulin Interaction
and analyzing the Biacore data. A.B. acknowledges the
support of a MBRS fellowship and an individual NRSA
predoctoral fellowship from the NIH. This work was
supported by NIH grant HL47463 to EAK and by NIH
grant HL62565 to A.R., and by the American Heart
Association.
References
Bode, W., Turk, D. & Karshikov, A. J. (1992). The
Ê crystal structure of d-Phe-Pro-Arg
re®ned 1.9 A
chloromethylketone-inhibited human a-thrombin:
structure analysis, overall structure, electrostatic
properties, detailed active site geometry, and structure-function relationships. Protein Sci. 1, 426-471.
Esmon, C. T. & Fukudome, K. (1995). Cellular regulation
of the protein C pathway. Semin. Cell Biology, 6,
259-268.
He, X., Ye, J., Esmon, C. T. & Rezaie, A. R. (1997). In¯uence of arginines 93, 97, and 101 of thrombin to its
functional speci®city. Biochemistry, 36, 8969-8976.
Janin, J. (1997). The kinetics of protein-protein recognition. Proteins: Struct. Funct. Genet. 28, 153-161.
Mandell, J. G., Falick, A. M. & Komives, E. A. (1998).
Identi®cation of protein-protein interfaces by
decreased amide proton solvent accessibility. Proc.
Natl Acad. Sci. USA, 95, 14705-14710.
Mathews, I. I., Padmanabhan, K. P., Tulinsky, A. &
Sadler, J. E. (1994). Structure of a non-adecapeptide
of the ®fth EGF domain of thrombomodulin complexed with thrombin. Biochemistry, 33, 13547-13552.
McCammon, J. A. (1998). Theory of biomolecular recognition. Curr. Opin. Struct. Biol. 8, 245-249.
Ni, F., Konishi, Y. & Scheraga, H. A. (1990). Thrombinbound conformation of the C-terminal fragments of
hirudin determined by transferred nuclear Overhauser effects. Biochemistry, 29, 4479-4489.
Nicholls, A., Sharp, K. A. & Honig, B. (1991). Protein
folding and association: insights from the interfacial
and thermodynamic properties of hydrocarbons.
Proteins: Struct. Funct. Genet. 11, 281-296.
Pearce, K. H., Ultsch, M. H., Kelley, R. F., de Vos, A. M.
& Wells, J. A. (1996). Structural and mutational
analysis of af®nity-inert contact residues at the
growth hormone-receptor interface. Biochemistry, 35,
10300-10307.
Perrett, S., Zahn, R., Stenberg, G. & Fersht, A. R. (1997).
Importance of electrostatic interactions in the rapid
binding of polypeptides to GroEL. J. Mol. Biol. 269,
892-901.
Radic, Z., Kirchhoff, P. D., Quinn, D. M., McCammon,
J. A. & Taylor, P. (1998). Electrostatic in¯uence on
the kinetics of ligand binding to acetylcholinesterase. J. Biol. Chem. 272, 23265-23277.
Raghavan, M. & Bjorkman, P. J. (1995). Biacore: a microchip-based system for analyzing the formation of
macromolecular complexes. Structure, 3, 331-333.
Raman, C. S., Jemmerson, R., Nall, B. T. & Allen, M. J.
(1992). Diffusion-limited rates for monoclonal antibody binding to cytochrome c. Biochemistry, 31,
10370-10379.
Schreiber, G. & Fersht, A. R. (1993). Interaction of Barnase with its polypeptide inhibitor barstar studied
by protein engineering. Biochemistry, 32, 5145-5150.
Schreiber, G. & Fersht, A. R. (1996). Mechanism of
rapid, electrostatically assisted, association of proteins. Nature Struct. Biol. 3, 427-431.
Schuck, P. & Minton, A. P. (1996). Kinetic analysis of
biosensor data: elementary tests for self-consistency.
Trends Biochem. Sci. 21, 458-460.
Szabo, A., Stolz, L. & Granzow, R. (1995). Surface plasmon resonance and its use in biomolecular interaction analysis (BIA). Current Biology, 5, 699-705.
Tsiang, M., Lentz, S. R. & Sadler, J. E. (1992). Functional
domains of membrane-bound human thrombomodulin. J. Biol. Chem. 267, 6164-6170.
Tsiang, M., Jain, A. K., Dunn, K. E., Rojas, M. E., Leung,
L. L. & Gibbs, C. S. (1995). Functional mapping of
the surface residues of human thrombin. J. Biol.
Chem. 270, 16854-16863.
Vindigni, A., White, C. E., Komives, E. A. & Di Cera, E.
(1997). Energetics of thrombin-thrombomodulin
interaction. Biochemistry, 36, 6674-6681.
Wells, C. M. & Di Cera, E. (1992). Thrombin is a Na(‡)activated enzyme. Biochemistry, 31, 11721-11730.
White, C. E., Hunter, M. J., Meininger, D. P., White, L. R.
& Komives, E. A. (1995). Large scale expression,
puri®cation and characterization of the smallest
active fragment of thrombomodulin: the roles of the
sixth domain and of methionine-388. Protein Eng. 8,
1177-1187.
Wu, Q. Y., Sheehan, J. P., Tsiang, M., Lentz, S. R.,
Birktoft, J. J. & Sadler, J. E. (1991). Single amino
acid substitutions dissociate ®brinogen-clotting and
thrombomodulin-binding activities of human
thrombin. Proc. Natl Acad. Sci. USA, 88, 6775-6779.
Xavier, K. A. & Wilson, R. C. (1998). Association and
dissociation kinetics of anti-hen egg lysozyme
monoclonal antibodies HyHEL-5 and HyHEL-10.
Biophys. J. 74, 2036-2045.
Xavier, K. A., McDonald, S. M., McCammon, J. A. &
Willson, R. C. (1999). Association and dissociation
kinetics of bobwhite quail lysozyme with monoclonal antibody HyHEL-5. Protein Eng. 12, 79-83.
Edited by J. A. Wells
(Received 27 May 1999; received in revised form 8 December 1999; accepted 8 December 1999)