You are on page 1of 12

Force dependent transition rates in chemical kinetics models for motor proteins

Gianluca Lattanzi and Amos Maritan

Citation: The Journal of Chemical Physics 117, 10339 (2002); doi: 10.1063/1.1521422
View online: http://dx.doi.org/10.1063/1.1521422
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/117/22?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in


Effect of gatekeepers on the early folding kinetics of a model β-barrel protein
J. Chem. Phys. 119, 5722 (2003); 10.1063/1.1599281

On the interpretation of force extension curves of single protein molecules


J. Chem. Phys. 116, 7760 (2002); 10.1063/1.1466835

Gaussian model of protein folding


J. Chem. Phys. 112, 1050 (2000); 10.1063/1.480629

On the transition coordinate for protein folding


J. Chem. Phys. 108, 334 (1998); 10.1063/1.475393

Characterization of foldable protein models: Thermodynamics, folding kinetics and force field
J. Chem. Phys. 107, 8089 (1997); 10.1063/1.475072

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
JOURNAL OF CHEMICAL PHYSICS VOLUME 117, NUMBER 22 8 DECEMBER 2002

Force dependent transition rates in chemical kinetics models


for motor proteins
Gianluca Lattanzi
Hahn-Meitner Institut, Abt. Theoretische Physik SF5, Glienickerstrasse 100, 14109 Berlin, Germany
Amos Maritan
International School for Advanced Studies (S.I.S.S.A.) and INFM, Via Beirut 2-4, 34014 Trieste, Italy
and The Abdus Salam International Center for Theoretical Physics, Strada Costiera 11, 34100 Trieste, Italy
共Received 15 July 2002; accepted 20 September 2002兲
We analyze the role of external forces 共both chemical and mechanical兲 in the kinetics of motor
proteins. Based on a generalized detailed balance condition, simple exponential force dependent
transition rates are widely used to interpret the available data. Yet, the use of Fokker–Planck
equations in continuous models allows for a direct insertion of the force. We describe an analytical
approach, based on a renormalization group scheme, to calculate the force dependence of transition
rates in a generic model. Our analysis shows that the simple exponential is a good approximation to
the correct force dependence only at low values of forces and provided that the step sizes are very
small. The law for the force dependent transition rates is tested on a set of data on kinesin, obtaining
a good agreement with existing results and predictions for future experiments. © 2002 American
Institute of Physics. 关DOI: 10.1063/1.1521422兴

I. INTRODUCTION size. Furthermore they are more likely to be generalized, so


that complicated chemical pathways can be investigated and
Experiments performed at the single molecule level are theoretical predictions can be compared directly to experi-
revolutionizing our understanding of motor proteins, myo- mental results 共see Kolomeisky and Fisher,3 and references
sins, kinesins, RNA polymerases, and many others. These therein for examples兲. However the load dependence of tran-
proteins are able to transduce the chemical energy of ATP sition rates in discrete models is still an open question. It is
hydrolysis into mechanical work, that is used for many dif- generally believed that a tangential force F in the opposite
ferent purposes, ranging from muscle contraction to transport direction of motion of a diffusing particle would decrease its
of material in nerve cells. The results of these experiments forward transition rate, k ⫹ , for a step ␦, by the appropriate
are often difficult to interpret, yet they constitute the best Arrhenius factor,

冉 冊
source of information on the kinetic mechanism at the basis
of this important process. Stochastic models are particularly F␦
suited to the interpretation of data that have been accumu- k ⫹ 共 F 兲 ⫽k ⫹ 共 0 兲 exp ⫺ . 共1兲
k BT
lated in an impressive fashion in the last few years. In par-
ticular molecular force clamps have made it possible to exert The same reasoning would yield an exponential increase
a constant force on motor proteins, and measure their veloci- of the backward step with the same applied load. This in turn
ties and diffusion constants under different loads, during leads to the prediction of exponential force dependence of
their normal cycle of operation. The effect of load on chemi- velocity, which is clearly unphysical. However a FP formu-
cal kinetics is an intriguing open question in current biologi- lation of the problem would imply a linear increase of veloc-
cal research and a fascinating field of interest for theorists in ity with force.
statistical physics and probabilistic methods. Indeed continu- In this paper we develop an analytical approach, based
ous models based on a system of coupled Fokker–Planck on a renormalization group scheme, to calculate the force
共FP兲 equations1,2 provide a physically based answer to the dependence of transition rates in discrete models. The ap-
question of the effect of a force on the kinetics of motor proach is quite simple: In the limit of very small displace-
proteins and mechanochemical systems in general. Yet these ments, discrete models must be equivalent to continuous
models are not directly suitable for the interpretation of ex- ones. This ‘‘continuum limit’’ has been already discussed.4
perimental data, since the calculations involved are rather Once the transition rates have been established in the con-
cumbersome, the connection to the actual chemical cycle is tinuum limit, a renormalization group scheme is used to
rather obscure, and, more importantly, they have been used eliminate a number of microscopic states of the system in a
so far solely on infinitely periodic systems, while most of the self-consistent way. After renormalization, we are left with a
experiments are usually performed on a finite size scale. simple formula for the force dependence of transition rates in
On the other hand, chemical kinetics or discrete models chemical kinetics models. This formula is tested on a set of
are more flexible, since they present the clear advantage of data for the motor protein kinesin, leading to the interpreta-
rather simple calculations, a direct connection to the actual tion of current results, predictions for future experiments and
chemical cycle and the applicability to systems of any finite hints on the kinetics of the energy transduction process.

0021-9606/2002/117(22)/10339/11/$19.00 10339 © 2002 American Institute of Physics


This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
10340 J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 G. Lattanzi and A. Maritan

II. THE GENERAL MODEL mechanical force is coupled to the position of the motor
protein along the filament, while the chemical force, i.e., the
We refer to the general picture proposed in a previous chemical potential difference in the hydrolysis reaction, is
paper.4 The system is described as a collection of states in a coupled to the number of ATP molecules in solution.
periodic two-dimensional manifold in which one direction We assume that transitions are possible only between
represents the position of the center of mass along the linear neighboring states, along both mechanical and chemical di-
track 共the microtubule or the actin filament兲, whereas the rections. Of course, this is a simplification; more general
other is the reaction coordinate for the ATP hydrolysis.5 The schemes with parallel branches and/or deaths can be found in
probability of being in a particular state-X at time t is written literature.3
as P X (t). The quantity W XY is defined as the transition prob- For a more detailed analysis and a direct comparison
ability per unit time from state Y to state X. It is assumed to with experiments on kinesin, we refer to the simpler version
be time independent. of the one-dimensional hopping models, already introduced.4
The time evolution of the system is simply given by the Such a simplification is appropriate for kinesin. Indeed the
master equation, mechanism for the energy transduction process has not been
resolved in motility assays: it may well be a power stroke
Ṗ X ⫽ 兺Y 共 W X,Y P Y ⫺W Y ,X P X 兲 . 共2兲 mechanism or a rectified Brownian movement, or even a
combination of the two. As observed by Fox and Choi,7 both
Since the system is cyclic, transition rates are assumed to mechanisms can be incorporated equally well into the gen-
be periodic, eral continuous model proposed by Keller and Bustamante,5
W X⫹LN,Y ⫹LN ⫽W X,Y , 共3兲 of which our model represents a discretized version.
We simplify the notation introducing a succession of N
where LN⬅(l 1 N 1 ,l 2 N 2 ), N 1 , N 2 are the periodicities in the states in one mechanochemical cycle and transition rates k ⫾i
mechanical and chemical direction, respectively, and l 1 , l 2 in the positive 共negative兲 direction of motion from state i (i
are integers. For convenience, we have also introduced the ⫹1) to state i⫹1 (i). We also assume that chemical energy,
probabilities and transition rates summed over all periods, expressed by the chemical potential difference ⌬ ␮ , is used
following the formalism of Derrida,6 to overcome the first barrier in the transition from state 1 to
state 2 共see our previous work4 for a discussion on this
R X⬅ 兺L P X⫹LN , 共4兲 point兲.
The next essential step is the choice of the potential of
mean force. Of course many choices are possible and, in the
WX,Y ⬅ 兺L W X,Y ⫹LN . 共5兲 absence of exact calculations or direct measurements,
equally suited. We adopt the simplest potential, with the
The time evolution of R X is easily obtained from Eqs. minimum number of parameters, which allows for an opti-
共2兲–共5兲, mal fit of the available experimental data; it may reveal the
essential steps in the mechanism of energy transduction and,
Ṙ X ⫽ 兺Y ⬘共 WX,Y R Y ⫺WY ,X R X 兲 , 共6兲 at least, can be invalidated easily by future experiments.
Although GDB is sufficient to recover the FP description
where, by definition, a primed sum is restricted only to one in the continuum limit,4 it is not enough to determine the
period along any axis. force dependence of mesoscopic transition rates, which are
The main advantage of this approach is its direct con- currently used in discrete chemical kinetics models. A sen-
nection to the continuous stochastic models, where the exter- sible coarse graining procedure, described in the subsequent
nal force term is easily inserted into the equations. As sections, will allow us to determine this dependence starting
pointed out in our previous work,4 a simple generalized de- from a continuous formulation.
tailed balance condition 共GDB兲 is fully consistent with the
requirement that, in the continuum limit, a FP equation is III. COARSE GRAINING PROCEDURE
recovered. We therefore require that transition rates satisfy
the GDB, We begin with the one-dimensional version of the model
described by Eq. 共6兲. The time evolution of R x , defined in
WX,Y 共 F 兲 R̂ Y 共 0 兲 e F•Y ⫽WY ,X 共 F 兲 R̂ X 共 0 兲 e F•X , 共7兲 Eq. 共4兲, is given by the following master equation:
where R̂ X (0) is the 共unique兲 stationary equilibrium solution Ṙ x ⫽Wx,x⫹1 R x⫹1 ⫹Wx,x⫺1 R x⫺1
of Eq. 共6兲 at F⫽0 and WX,Y (F) is the force dependent tran-
sition rate, still unknown. ⫺ 共 Wx⫹1,x ⫹Wx⫺1,x 兲 R x , 共8兲
It is important at this point to remark what we mean by where we are considering only a set of N discrete spatial
external force. In the GDB framework we use a generalized positions. The step performed by the motor in one transition
force, defined as an externally tuned parameter able to drive is given by ⌬x⫽1/N, where the period has been chosen as
the system out of equilibrium. Any direction in the two- the unit length. We introduce the Laplace transform,
dimensional manifold we are considering, entails a general-
ized force along that direction that is coupled, via GDB, to
the state variables along the same direction. For instance the
R̃ x 共 s 兲 ⫽ 冕
0
⫹⬁
exp共 ⫺st 兲 R x 共 t 兲 . 共9兲
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 Kinetics models for motor proteins 10341

The W dependence of R̃ x (s) is not indicated for simplic- where f is an applied external force, k B is the Boltzmann
ity. The stationary condition is obtained in the limit t→⬁; constant, and T is the temperature. We assume the potential
this limit is equivalent to the limit s→0 for the Laplace of mean force not to be affected by the presence of an exter-
transform. Applying this transformation to both sides of Eq. nal force, i.e., that the force is not able to alter the confor-
共8兲, we find mation of the protein. Different potentials may account also
for the case of different protein conformations; nonetheless
共 s⫹Wx 兲 R̃ x ⫽R x 共 t⫽0 兲 ⫹Wx,x⫹1 R̃ x⫹1 ⫹Wx,x⫺1 R̃ x⫺1 , we will restrict our study to the simplest case. Equation 共14兲
共10兲 implies that the new transition rates satisfy the same Eq. 共17兲
where we have defined Wx ⫽Wx⫹1,x ⫹Wx⫺1,x . Since we are with the new lattice spacing ⌬x ⬘ ,

冉 冊
interested only in the stationary solution, which is unique4,8
and reached independently of the initial condition, we as- ⬘
Wx⫹1,x ⫺V 关 2 共 x⫹1 兲兴 ⫹V 共 2x 兲 ⫹ f ⌬x ⬘
⫽exp . 共18兲
sume the initial condition to be R x (t⫽0)⫽ ␦ x,0 . ⬘
Wx,x⫹1 k BT
Now we introduce a decimation procedure over odd
sites:9 the idea is that the system properties should not de- Thus assuming that the GDB holds for the microscopic
pend on the particular choice of ⌬x. Therefore with our deci- transition rates, the GDB is preserved under the decimation
mation procedure we double the step ⌬x, and we ought to procedure, and hence, holds also for the mesoscopic transi-
obtain an evolution equation with the same form as Eq. 共10兲 tion rates, obtained after many iterations of the decimation
procedure. This simply means that the ratio of the forward
with new transition rates W⬘x ⬘ ,y ⬘ with x ⬘ ⫽x/2, y ⬘ ⫽y/2, and
and reverse rate constants gives the correct force depen-
⌬x ⬘ ⫽2⌬x. To this purpose, we just need to solve Eq. 共10兲
dence, as already reported,10,11 even though the single rate
for R̃ 2x⫹1 ,
constants are not Arrhenius-type. This property enforces our
W2x⫹1,2共 x⫹1 兲 W2x⫹1,2x confidence in the coarse graining procedure.
R̃ 2x⫹1 ⫽ R̃ 2 共 x⫹1 兲 ⫹ R̃ 2x . 共11兲
s⫹W2x⫹1 s⫹W2x⫹1 IV. APPLICATION TO LINEAR POTENTIALS
Writing the same equation for R̃ 2x⫺1 and substituting We apply the coarse graining procedure to a potential
these expressions in the equation for R̃ 2x , we finally obtain which is linear in an interval 关 0,L 兴 ; this example is of great
R̃ ⬘x 共 s ⬘ ⫹Wx⬘ 兲 ⫽ ␦ x,0⫹Wx,x⫹1
⬘ ⬘ ⫹Wx,x⫺1
R̃ x⫹1 ⬘ ⬘ ,
R̃ x⫺1 共12兲 importance for two reasons: first, any potential shape can be
always approximated by a piecewise linear potential, second
where an eventual external force may be also described as a tilt in
the potential shape, and therefore by a linear potential.
R̃ x⬘ 共 s ⬘ 兲 ⫽A ⫺1 R̃ 2x 共 s 兲 , 共13兲
We therefore assume V(x) to be linear in x (dV/dx
W2x,2x⫾1 W2x⫾1,2共 x⫾1 兲 ⫽const), and we define

Wx,x⫾1 ⫽A , 共14兲
s⫹W2x⫾1 1 dV f
⫹ 共19兲

冋 册
F⫽⫺ .
W2x⫹ ⑀ ,2x k B T dx k B T
s ⬘ ⫽As 1⫹ 兺
⑀ ⫽⫾1 s⫹W2x⫹ ⑀
. 共15兲 We now write

The amplitude A in the above equations is, at the mo- W⫹ ⫽Wx⫹1,x ⫽e F ⫹ , 共20兲
ment, arbitrary and it can be fixed appropriately depending W⫺ ⫽Wx,x⫹1 ⫽e F ⫺ , 共21兲
on the purposes. In what follows, we choose it to be inde-
pendent of the iteration step 共other instances are possible9兲. F ⫹ ⫺F ⫺ ⫽F⌬x. 共22兲
Since R̃ ⬘x (s ⬘ ) is the solution to Eq. 共12兲, which has the same The last equations ensure the GDB for microscopic tran-
form as Eq. 共10兲, the functional dependence of R̃ x⬘ on s ⬘ and sition rates. We apply now the coarse graining procedure and
W⬘ is the same as that of R̃ x on s and W; therefore we can obtain in the s→0 limit,
safely omit the prime and use R̃ x (s ⬘ ) in subsequent calcula-
tions. AW⫹
⬘⫽
W⫹ , 共23兲
In the large time limit, e.g., in the s→0 limit, Eq. 共15兲 1⫹r
simplifies and becomes

冉 冊
ArW⫺
W2x⫹ ⑀ ,2x ⬘⫽
W⫺ , 共24兲
s ⬘ ⫽As 1⫹ 兺
⑀ ⫽⫾1 W2x⫹ ⑀
⫹o 关 s 2 兴 . 共16兲 1⫹r
with r⫽W⫺ /W⫹ ⫽e ⫺F⌬x . Also r is transformed in the
Now, let us assume that the generalized detailed balance, coarse graining procedure, so that r ⬘ ⫽r 2 . We iterate this
Eq. 共7兲, holds for the transition rates W’s; since R̃ x (0) procedure for n steps,
⬀exp(⫺ ␤ V(x)), where V(x) is the 共periodic兲 potential of n
mean force, we obtain r 共 n 兲 ⫽r 2 , 共25兲

Wx⫹1,x
Wx,x⫹1
⫽exp 冉 ⫺V 共 x⫹1 兲 ⫹V 共 x 兲 ⫹ f ⌬x
k BT
冊 , 共17兲 W共⫹n 兲 ⫽
兿 k⫽0
n⫺1
A n W⫹
共 1⫹r 共 k 兲 兲
, 共26兲

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
10342 J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 G. Lattanzi and A. Maritan

where n is such that


L⫽⌬x 共 n 兲 ⫽2 n ⌬x. 共27兲
2 n ⫺1 l
It is simple to show that 兿 k⫽0
n⫺1
(1⫹r (k) )⫽ 兺 l⫽0 r
n
⫽(r 2 ⫺1)/(r⫺1), so that
W⫹ 共 L,F 兲 ⬅W共⫹n 兲

⫽ 冉 冊
L
⌬x
共 ln A/ln 2兲
F⌬x
1⫺exp共 ⫺FL 兲
W⫹ 共 ⌬x 兲 ,

共28兲
in the small ⌬x limit. We assume that in this limit, the tran-
sition probability W⫹ ⬅W⫹ (⌬x) does not depend on the de- FIG. 1. Approximation of a generic potential of mean force with a piece-
tailed potential shape.28 As already shown,4 the FP equation wise linear potential, for which our coarse graining procedure can be per-
formed, leading to a 4 states model. Units are arbitrary.
is recovered in the continuum limit, when
D
W⫹ ⫽ , 共29兲 In motility assays, measurable quantities are usually
⌬x 2
given in terms of the load applied to the motor proteins,
where D is the local diffusion coefficient. Substituting Eq. which is simply a negative force in our model, ⫺ f .
共29兲 in Eq. 共28兲, we obtain A nice feature of Eq. 共31兲 is that no extra parameters are
introduced to account for a difference between forward and
F⌬x ⫺ 共 1⫹ln2 A 兲
W⫹ 共 L,F 兲 ⫽DL ln2 A . 共30兲 backward transition rates. As remarked by Astumian and
1⫺exp共 ⫺FL 兲 Bier,12 there is a priori no reason why we should use differ-
ent apportionment of forces between forward and backward
If we require that W⫹ does not depend on the particular
transition rates.
choice of ⌬x, the parameter A should take the value 1/2.
Our approach shows that a different apportionment of
With this substitution, we finally obtain our force dependent
force in a coarse grained model may arise even from an
transition rates,
unbiased microscopic mechanism, without invoking different
F exp共 ⫾FL/2兲 exponential prefactors. This, in turn, reduces the total num-
W⫾ 共 L,F 兲 ⫽D . 共31兲
2L sinh共 FL/2兲 ber of parameters needed to fit the experimental data.
A similar approach based on numerical integrations of a
With this choice we get also s ⬘ ⫽s⫹O(s 2 ) from Eq.
piecewise linear potential was used by Keller and
共15兲. Equation 共31兲 has been obtained under the following
Bustamante,5 to derive force dependent transition rates. De-
hypotheses:
tailed calculations indeed showed different behaviors in the
共i兲 The motion of the protein from one site to the other is forward and backward transitions. The same model predicted
the result of Brownian diffusion in a linear potential. forward 共backward兲 transition rates linear in force at large
共ii兲 A discrete chemical kinetics model with few states is positive 共negative兲 loads. This is clearly in contrast with
apt to describe this motion. models with pure exponential 共or Arrhenius兲 load depen-
共iii兲 The continuous FP description is recovered when the dence. As observed,5 this should be taken into account when
steps taken by the motor protein are small. fitting the experimental data with a pure exponential law, of
the Arrhenius form, since a fit of the rate constants to an
These hypotheses are rather general and could hence be exponential function would lead eventually to different val-
supposed to hold for a wide class of motor proteins and ues of the evaluated activation free energy from one range of
biological systems. forces to the other.
It is striking to observe that our model, based on a dif-
ferent approach, leads essentially to the same behavior,5 with
V. FORCE DEPENDENT TRANSITION RATES forward 共backward兲 rate constants linear at large negative
共positive兲 loads 共see Fig. 2兲 and almost exponential at small
We have applied our coarse graining to a linear potential. forces. Furthermore, models with pure exponential load de-
A more general potential of mean force can be approximated pendence would indeed predict positive exponential veloci-
by a piecewise linear potential as shown in Fig. 1. Our for- ties for high forces and negative exponential velocities for
mula can be applied to each of the linear pieces, and there- high loads, which is clearly unphysical. In our case the force
fore a discrete model is generated.29 For a potential barrier dependence is at most linear with high positive 共or negative兲
U, our model would predict force dependent forward/ forces.
backward rates of the form, Non-Arrhenius force dependence of the very same form
proposed in Eq. 共31兲 has been already reported for the total
D f L⫺U e ⫾ 共 f L⫺U 兲 / 共 2k B T 兲 velocity of a motor protein,13 simply assuming the GDB and
W⫾ ⫽ . 共32兲
k BT 2L 2 sinh关共 f L⫺U 兲 /2k B T 兴 a velocity linear with force at low loads.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 Kinetics models for motor proteins 10343

tiplied by the concentration of the chemical species involved


and a new parameter is introduced. The discussion reported
by Keller and Bustamante5 applies exactly to our model: if
we introduce a second order step to describe the mecha-
nochemical binding of ATP (T) to the motor protein (M ),
k 1共 f 兲
M ⫹T
M T, 共33兲
k ⫺1 共 f 兲

where MT represents the bound state of the motor protein,


the force dependence of reaction rates k 1 ( f ), k ⫺1 ( f ) implies
that the binding step involves net motion. Indeed, as con-
firmed by experimental results,14 and already observed,15 the
spatial step covered by the motor upon binding, should be
very small. Keller and Bustamante5 point out that every
FIG. 2. Forward–backward transition probability vs load. Free energy dif-
ference U/k B T⫽5, T⫽300 K, L⫽4 nm, and D⫽0.1 ␮m2/s.
binding process should involve at least two parts: a purely
second-order process in which two molecules come into
loose contact by diffusion alone, and a first order process in
The Arrhenius form is a good approximation only if all which the two molecules undergo conformational changes
mechanical substeps taken by the motor protein along the that result in a more strongly bound state. It is therefore
linear track are very small 关a reasonable choice of parameters possible to divide the single step above into an equivalent
in Eq. 共31兲 would indicate 1 nm substeps and 1 pN forces兴. two-step process,
Therefore, the purely exponential load dependence is a good
k1 k 2共 f 兲
approximation for chemical steps which are thought to occur
M 1 ⫹T
M 2 T
M 3 T, 共34兲
without a net advancement of the motor protein,14 but not for k ⫺1 k ⫺2 共 f 兲
all substeps and not for the whole range of forces studied in
motility assays. where M 1 , M 2 , M 3 represent different states of the motor
protein. The rate constants for the first step are assumed to be
VI. MINIMAL MODELS independent of force. This assumption means that the exter-
The general classification introduced by Keller and nal force acts on the protein, but not on the fuel molecule
Bustamante5 can be transferred to our model, with a great directly, so that the purely diffusive part of the binding is
simplification of the calculations and the number of param- force independent. In a minimal model, we may assume that
eters used. There are essentially 2N⫹1 parameters: the size a third step corresponding to ADP (D) dissociation, closes
of individual steps (N), the potential differences between the cycle, so that the motor is finally in its initial state M 1
successive positions (N) and an overall multiplicative factor and a new ATP molecule can bind. Furthermore we assume
related to the macroscopic diffusion coefficient. Two other this step to be described by two transitions: a first-order one,
parameters may be conveniently used to describe the chem- corresponding to conformational changes and a second one
istry of the reaction, i.e., the dependence of rate constants on accounting for release of products. Two choices are possible:
ATP concentration, and on reaction products. The total num- 共a兲 the conformational change is followed by the release of
ber of parameters is therefore 2N⫹3. Yet, two conditions products,
reduce the number of parameters: the sum of individual steps
k1 k 2共 f 兲 k 3共 f 兲 k4
is equal to one period, and the sum of potential differences
M 1 ⫹T
M 2 T
M 3 T
M 4 T
M 1 ⫹D; 共35兲
must yield zero since the system is periodic. The total is k ⫺1 k ⫺2 共 f 兲 k ⫺3 共 f 兲 k ⫺4
2N⫹1, to be compared with the 4N⫺1 parameters of mod-
els with Arrhenius dependence and different force apportion- 共b兲 the conformational change follows the release of prod-
ments. ucts. In line with the previous reasoning, we assume the con-
Our approach is therefore suited to generate a class of formational change to be force-dependent and the release of
minimal models, for which few parameters are needed. How- products to correspond to a negligible advancement of the
ever experimental results may suggest to enrich these mini- motor,
mal models with branch or death processes, or waiting time
k1 k 2共 f 兲 k3 k 4共 f 兲
distributions, for which detailed calculations have been re-
M 1 ⫹T
M 2 T
M 3 T
M 4 ⫹D
M 1 . 共36兲
cently performed.3 k ⫺1 k ⫺2 共 f 兲 k ⫺3 k ⫺4 共 f 兲
Our coarse graining procedure has been shown to be
consistent with the FP equation approach, in the continuum We observe that these models are more detailed than
limit, with a simplification of the mathematics involved. The those proposed by Keller and Bustamante.5 At least two tran-
obtained reaction rates depend only on the general shape of sition rates are force dependent. The detailed mechanochemi-
the potential, and not on its full details. The results above cal picture can be different and more complicated. We de-
apply only to first-order processes, but GDB accounts also cided to use these models since they represent a simple, yet
for second order processes. The reaction rate should be mul- general picture, consistent with the schemes proposed in the
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
10344 J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 G. Lattanzi and A. Maritan

FIG. 3. Simplified 4 state model for kinesin. ATP binding is used to over- FIG. 4. Simplified 4 state model for kinesin. ATP binding is used to over-
come the potential barrier U 1 and it is followed by a diffusive step of come the first potential barrier and it is followed by a diffusive step of
approximately 4 nm. A second barrier is overcome before product (D) re- approximately 4 nm and a drop in the potential of height U 1 . A small
lease with a second diffusive step of 4 nm. second barrier is overcome during product (D) release, which triggers a
second diffusive step of 4 nm.

recent literature.14,16 –18 The steps corresponding to ATP


binding and ADP dissociation are assumed to be purely dif-
to move towards the plus end of the microtubule, sliding
fusive, i.e., they are negligibly small in the range of force
down the first hill in the potential. Then a new conforma-
studied 共⫺6 pN, 6 pN兲. This approximation is valid as long
tional change triggers the release of reaction products and the
as f LⰆU.
attachment to the next microtubule binding site. The reduced
For other models in the same spirit, and for a compari-
potential difference for the second barrier accounts for the
son with experiments, the interested reader could refer to
difficulty to undergo this conformational change.
Mogilner et al.,18 where a formula with close resemblance to
Using this model we were not able to fit the data.21 It is
Eq. 共31兲 is derived from simple arguments for the force de-
indeed possible to show that the Michaelis constant for such
pendence of some transition rates.
a model depends linearly on the maximum attained velocity,
therefore it decreases with load as long as the maximum
VII. CHOICE OF POTENTIAL attained velocity does so.
We therefore switch to model 共b兲. The potential for this
The advances in the field of single molecule manipula- model is represented in Fig. 4. In this case we assume that
tion for kinesin have permitted to measure the velocity under the product release takes place when the protein is in a local
an externally applied load at a given ATP concentration. At minimum at a distance of 4 nm from both microtubule bind-
the same time it is possible to measure the randomness pa- ing sites. The release of products leads to a second confor-
rameter, which is directly related to the macroscopic diffu- mational change which finally leaves the protein in its initial
sion constant. In our previous work,4 we showed how to state. We will show here only results for this simplified
calculate the velocity for any discrete model, whereas the 4-state mechanochemical reaction model 共b兲, Eq. 共36兲.
macroscopic diffusion constant was calculated only for The master equation is given by the following system of
simple 2- and 3-state models. We sketch in the Appendix equations:5
how to calculate the diffusion constant in any discrete model,
with the formalism already introduced, using the method Ṙ 1 ⫽⫺ 共 k ⫺4 ⫹k 1 关 T 兴 兲 R 1 ⫹k ⫺1 R 2 ⫹k 4 R 4 ,
proposed by Koza.19,20
We now need to give an estimate for all the parameters Ṙ 2 ⫽k 1 关 T 兴 R 1 ⫺ 共 k ⫺1 ⫹k 2 兲 R 2 ⫹k ⫺2 R 3 ,
of the model, in particular for the potentials. For model 共a兲
Ṙ 3 ⫽k 2 R 2 ⫺ 共 k ⫺2 ⫹k 3 兲 R 3 ⫹k ⫺3 关 D 兴 R 4 , 共37兲
we assume the simple potential depicted in Fig. 3. The
chemical energy of ATP binding is used to overcome the first Ṙ 4 ⫽k ⫺4 R 1 ⫹k 3 R 3 ⫺ 共 k ⫺3 关 D 兴 ⫹k 4 兲 R 4 ,
barrier in the potential U 1 , then energy is released, when the
motor head diffuses in the positive direction, so that a sub- where 关 T 兴 and 关 D 兴 represent the ATP and product 共ADP and
step of ⬃4 nm is performed. The reason for choosing sub- P i ) concentrations in the two second-order steps involved.
steps of this size resides in the experiments reported by the The stationary solution to Eq. 共37兲 allows us to compute all
Yanagida’s lab.17 quantities of interest, as shown for a general class of discrete
As assumed above, the product release does not corre- models.4,5 The velocity is simply given by the filament peri-
spond to any net advancement of the motor protein, and it is odicity, p⫽8 nm in the case of kinesin, times the next flux in
only driven by a difference in chemical potential, so that a the diffusive steps, M 2 T
M 3 T and M 4
M 1 in our case,
second minimum in the potential may arise. Although the a 关 T 兴 ⫺b 关 D 兴
kinetic mechanism leading to this potential is rather obscure, v⫽p , 共38兲
e 关 T 兴 ⫹ f 关 T 兴关 D 兴 ⫹g 关 D 兴 ⫹h
our picture is that ATP triggers a conformational change
which is responsible for the likelihood of the center of mass with
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 Kinetics models for motor proteins 10345

TABLE I. Transition rates for the 4 states model described by Eq. 共36兲. The force, f, must be expressed in pN,
the ATP concentration, 关 T 兴 , in ␮M. The model corresponds to two substeps, each of size 4 nm; the potential
constants appearing in Fig. 4 are U 2 ⫽12 k B T and U 4 ⫽4 k B T at room temperature, T⫽298 K.

k1 3 ␮M⫺1 s⫺1 k ⫺1 3000 s⫺1


共 6⫹0.48• f 兲 •e 6⫹0.48• f
共 6⫹0.48• f 兲 •e ⫺6⫺0.48• f
k2 200 s⫺1 k ⫺2 200 s⫺1
sinh共 6⫹0.48• f 兲 sinh共 6⫹0.48• f 兲

k3 140 s⫺1 k ⫺3 undetermined

共 2⫹0.48• f 兲 •e 2⫹0.48• f 共 2⫹0.48• f 兲 •e ⫺2⫺0.48• f


k4 200 s⫺1 k ⫺4 200 s⫺1
sinh共 2⫹0.48• f 兲 sinh共 2⫹0.48• f 兲

a⫽k 1 k 2 k 3 k 4 , 关ATP兴 at small concentrations, while it saturates at v max for


high concentrations. The quality of these fits can be judged
b⫽k ⫺1 k ⫺2 k ⫺3 k ⫺4 , from a direct comparison with experimental data.21 Our
e⫽k 1 共 k 2 k 3 ⫹k ⫺2 k 4 ⫹k 2 k 4 ⫹k 3 k 4 兲 , model predicts a drop in velocity at about 6 –7 pN, the same
stall load observed in experiments. It is interesting to observe
f ⫽k 1 k ⫺3 共 k 2 ⫹k ⫺2 兲 , that our model predicts an exponential drop of velocity and
g⫽k ⫺3 共 k ⫺2 k ⫺4 ⫹k ⫺1 k ⫺4 ⫹k ⫺1 k ⫺2 ⫹k 2 k ⫺4 兲 , not an inversion of the direction of motion. The stall load can
be calculated, assuming that experimental devices are not
h⫽ 共 k ⫺1 k ⫺2 ⫹k ⫺1 k 3 ⫹k 2 k 3 兲共 k 4 ⫹k ⫺4 兲 . able to detect velocities below a certain cutoff 共dictated by
In the limit of small product concentrations 关 ADP兴 experimental precision兲. It is however important to notice
关 P i 兴 , transitions from state 3 to state 4 共ADP release兲 are
that this model does not predict an inversion of velocity; to
irreversible and the model is further simplified: the velocity our knowledge, no motor protein has been observed to re-
follows Michaelis–Menten kinetics, verse its velocity even under high forces: backward steps are
possible, but the sign of velocity is always the same. This is
v max关 T 兴 an important and strong prediction of our model.
v⫽ , 共39兲 In Fig. 6, we plot the velocity versus ATP concentra-
关 T 兴 ⫹K M
tions, at different loads. We observe that the velocity follows
where v max is the maximum attained velocity at high 关 T 兴 , the Michaelis Law, at all concentrations and all values of
force. We have already reported that the Michaelis Law
a k 2k 3k 4
v max⫽p ⫽ p , 共40兲 should be modified to account for the possibility of steps in
e k 2 k 3 ⫹k ⫺2 k 4 ⫹k 2 k 4 ⫹k 3 k 4 the negative direction of motion.15 Our results from the dis-
and K M is the Michaelis constant, crete model seem to be in contrast with those findings; in-
deed our discrete models do not allow backward cycles to be
h 共 k ⫺1 k ⫺2 ⫹k ⫺1 k 3 ⫹k 2 k 3 兲共 k 4 ⫹k ⫺4 兲 completed, since we assumed that the concentration of the
KM⫽ ⫽ . 共41兲
e k 1 共 k 2 k 3 ⫹k ⫺2 k 4 ⫹k 2 k 4 ⫹k 3 k 4 兲 reaction products is 0. In real experiments, this concentration
is assumed to be negligible, but it is not zero; if backward
Since some of the transition rates are force dependent, cycles are possible, a correction to the Michaelis Law in the
the Michaelis constant is force dependent, as experimentally
observed.

VIII. COMPARISON WITH EXPERIMENTAL RESULTS


The choice of parameters is a difficult problem, since no
information on the detailed mechanism is available. We de-
cided to use the reaction rates reported by Moyer et al.22 as
starting values, and then varied these parameters so as to fit
the data from Block’s lab.21,23 The number of parameters is 3
for the chemical steps (M 1 ⫹T
M 2 T and M 3 T
M 4
⫹D), and 5 for the force dependent ones 共the two step sizes,
the two potential differences and the multiplicative factor
which depends on the step size and therefore is the same for
both mechanical steps兲. All transition rates obtained by the
fit, are reported in Table I.
FIG. 5. Load-velocity curves for different ATP concentrations. Top to bot-
In Fig. 5, we plot the velocity attained by the motor tom curves: 关 T 兴 ⫽2 mM, 1 mM, 100 ␮M, and 5 ␮M. Filled circles refer to
protein under different loads, for different ATP concentra- experimental data from Visscher et al. 共Ref. 21兲 at 关 T 兴 ⫽2 mM and 5 ␮M.
tions. The zero load velocity is approximately linear with The size of the data points represents the experimental error.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
10346 J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 G. Lattanzi and A. Maritan

FIG. 6. Michaelis–Menten kinetics under load. Double logarithmic plot of


the average bead velocity, vs ATP concentration for various loads (⫺ f ). FIG. 8. Randomness parameter vs ATP concentrations at different 共positive
Experimental data are shown with different point types. and negative兲 loads. The curve at 0 load is shown as a solid line, for refer-
ence.

form already proposed, appears also in the model.


Remarkably, our model allows for a direct calculation of 具 x 2共 t 兲 典 ⫺ 具 x 共 t 兲 典 2
the Michaelis constant, K M , which is expressed by Eq. 共41兲. D⫽ lim . 共42兲
K M depends on force and is plotted in Fig. 7; it increases t→⬁ 2t
with load, as observed.21 The model not only agrees, quali- Full details of the calculations needed to estimate the
tatively and quantitatively, with experimental results, but it macroscopic diffusion coefficient are given in the Appendix.
also allows us to calculate K M in the region of negative The method is applied to a 4 state model, but it can be
loads. In particular we find that K M should show a much less generalized to models with any number of states using the
pronounced variation with negative loads and should slowly method introduced by Koza.19,20 Here we present the results
decrease with force. in terms of randomness, a dimensionless parameter,25 which
In our opinion, it is crucial to compare these results with is related to the diffusion coefficient,
experiments. To our knowledge, only an increase in the
Michaelis constant in the region of positive loads has been 2D
r⫽ , 共43兲
observed so far. The region of positive force has been pv
investigated,24 but the Michaelis constant and its variation where p is the filament periodicity and v is the motor veloc-
with force have not been measured yet. Therefore, new ex- ity. In the case of motor proteins, this parameter is more
periments aimed at investigating this parameter region, al- appropriate, since its reciprocal represents the number of rate
though difficult, are extremely important to confirm or reject limiting steps. See Wang et al.26 for a discussion on this
our predictions. point.
The macroscopic diffusion coefficient is defined as the We plot in Fig. 8 the randomness parameter as a function
long time limit of the ratio between the variance of the pro- of ATP concentration at different loads. Its variability with
tein position along the filament, 具 x 2 (t) 典 ⫺ 具 x(t) 典 2 , and time t, load is not trivial at all; indeed the agreement with the ex-
perimental data is only of a qualitative nature. We observe
that our model predicts a randomness parameter always
smaller than 1. This is in contrast with experiments, where
randomness parameters ⬎1 have been measured.21 As
pointed out in the same reference, the increase in r above
unity probably reflects the occurrence of 8-nm backward
steps. In our model, these backward steps are not allowed by
the choice of an irreversible reaction rate k 3 due to the lack
of reaction products. If the cycle is allowed to be reversible,
keeping a finite k ⫺3 and a not negligible product concentra-
tion, the randomness does increase above unity in some re-
gions of the parameter space, and remarkably in those re-
gions where the velocity is close to zero 共typically at low
ATP concentrations and large forces兲. Indeed, backward
steps are observed with a higher probability under high
loads,17 hence they are likely to raise the observed random-
FIG. 7. Michaelis constant at different loads. The predicted behavior is in
ness factor above our predicted value 0.5. Following these
agreement with experimental data 共filled circles in the figure兲: the Michaelis
constant increases with applied load. considerations, our predictions should be used only as esti-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 Kinetics models for motor proteins 10347

search for consistency between discrete models and continu-


ous ones. In this respect, we have shown that the load de-
pendence of transition rates in mechanochemical systems can
be more complicated than the Arrhenius exponential. We
have provided an analytical method, based on a renormaliza-
tion group scheme, leading to a new formula for the force
dependence of transition states. This formula has been used
to interpret the available data on the kinetics of the motor
protein kinesin, obtaining a remarkable agreement with ex-
perimental results and theoretical predictions for future ex-
periments. We are convinced that the analytical tools devel-
oped for the case of motor proteins, can be easily generalized
to other biological systems and can be useful in a number of
FIG. 9. Randomness factor measured for various loads at different ATP experiments at the single molecule level.
concentrations.

ACKNOWLEDGMENTS
mates on the minimum values for the randomness parameter
observed in real experiments. The randomness parameter is One of the authors’ 共G.L.兲 research has been supported
smaller than 1/2, indicating that more than two steps are by a Marie Curie Fellowship of the European Community
rate-limiting at low loads and low ATP concentrations. At Program ‘‘Improving Human Research Potential and the
high ATP concentrations and low loads, there are essentially Socio-Economic Knowledge Base’’ under Contract No.
two rate-limiting steps. Therefore the first step is rate limit- HPMF-CT-2001-01432. The authors are solely responsible
ing at low ATP concentrations, while at high ATP concentra- for information communicated and the European Commis-
tions, the third and fourth steps are rate limiting. Transition sion is not responsible for any view or results expressed.
rates for these steps are not numerically equivalent, hence the
predicted randomness factor ⬎1/2. At high loads and high
ATP concentrations, there are essentially two rate-limiting APPENDIX: DIFFUSION COEFFICIENT
steps, the third and fourth ones and they are almost equiva-
lent, r⫽1/2. The evaluation of the diffusion coefficient in periodic
We remark that the same crossing of curves has been systems is an interesting problem in the theory of stochastic
observed in experiments21 and that our results were obtained processes, and is also of great importance since experiments
by a simple fit to a small set of data for the maximum at- performed on motor proteins usually measure this coeffi-
tained velocity. In Fig. 8, we also plot the randomness in the cient, or the related randomness parameter. The approach
region of negative loads. Again results in this parameter re- used here is similar to the one proposed by Wang et al.26 for
gion, to our knowledge, are not yet available. RNA polymerase and here adapted to our 4 states model. The
In Fig. 9 we plot the randomness parameter as a function procedure has been generalized to models with N states,19,20
of the applied force at different ATP concentrations. Mea- and is a valid alternative to the calculations involved in dis-
surements at high loads are extremely difficult, since the mo- crete jump processes,6 especially when the discrete steps are
tor is likely to detach from the microtubule,30 as observed.21 different from each other, as is the case of our model.
An inspection of Fig. 9 reveals that at low ATP concentra- We can derive an expression for the effective diffusion
tions, there are essentially 2 rate-limiting steps, the first one coefficient by examining the evolution of a solution of the
form,

冋 册 冋 册
and the fourth one which becomes the only rate-limiting step
at high loads. At higher ATP concentrations, only one step is
P 1 共 n,t 兲 c 1共 t 兲
rate limiting, the force dependent one, as revealed by the 100
␮M curve. Indeed at low loads, the first step is still rate P 2 共 n,t 兲 c 2共 t 兲
⫽e iknp , 共A1兲
limiting at 100 ␮M ATP, but the numerical value of the cor- P 3 共 n,t 兲 c 3共 t 兲
responding transition rate is considerably higher than the one P 4 共 n,t 兲 c 4共 t 兲
corresponding to the force dependent step, hence the ran- where P i (n,t) represents the probability to be in state i, i
domness parameter approaches one from below at high ⫽1,2,3,4 in the period n, p is the periodicity of the filament,
loads. At rather high ATP concentrations 共1 and 2 mM curves c i (t) are unknown functions of time. This is the solution of
in Fig. 9兲, there are essentially two rate limiting steps, the Eq. 共6兲. Substituting Eq. 共A1兲 in Eq. 共6兲, we find the time
third and fourth ones, and finally only one at very high loads.

冋 册冋 册
evolution of the unknown c i (t),
c 1共 t 兲 c 1共 t 兲
IX. SUMMARY AND CONCLUSIONS d c 2共 t 兲 c 2共 t 兲
⫽A , 共A2兲
dt c 3共 t 兲 c 3共 t 兲
We have presented theoretical arguments aimed at the
c 4共 t 兲 c 4共 t 兲
identification of the role of applied forces on the kinetics of
motor proteins. The driving principle for our work is the where
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
10348 J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 G. Lattanzi and A. Maritan

冋 册
⫺ 共 k 1 关 T 兴 ⫹k ⫺4 兲 k ⫺1 0 k 4 exp共 ⫺ik p 兲
k 1关 T 兴 ⫺ 共 k ⫺1 ⫹k 2 兲 k ⫺2 0
A⫽ , 共A3兲
0 k2 ⫺ 共 k ⫺2 ⫹k 3 兲 0
k ⫺4 exp共 ik p 兲 0 k3 ⫺k 4

for model 共b兲. The eigenvalues of the matrix A are the roots D⫽ p 2 b. 共A12兲
of the characteristic polynomial,
For a detailed mathematical proof, the reader is referred
det共 A⫺␭I 兲 ⫽␭ 4 ⫹ ␴ 3 ␭ 3 ⫹ ␴ 2 ␭ 2 ⫹ ␴ 1 ␭⫹ ␴ 0 , 共A4兲 to the work by Koza.19,20 The advantage of such an approach
is that the characteristic polynomial, Eq. 共A4兲 is sufficient to
with
determine the coefficients a and b without detailed knowl-
␴ 3 ⫽k 1 关 T 兴 ⫹k 2 ⫹k 3 ⫹k 4 ⫹k ⫺1 ⫹k ⫺2 ⫹k ⫺4 , 共A5兲 edge of its roots.
Indeed a⫽ ⳵ ␭ 0 ( ⑀ )/ ⳵ ⑀ 兩 ⑀ ⫽0 . This coefficient can be
␴ 2 ⫽k 1 关 T 兴共 k 2 ⫹k ⫺2 ⫹k 3 ⫹k 4 兲 ⫹k 2 共 k 3 ⫹k 4 ⫹k ⫺4 兲 evaluated by writing ␭ 0 ( ⑀ ) in place of ␭ 0 in Eq. 共A4兲 and
⫹k 3 共 k 4 ⫹k ⫺4 兲 ⫹k ⫺1 共 k ⫺2 ⫹k 3 ⫹k 4 ⫹k ⫺4 兲 deriving both sides with respect to ⑀. This derivative is zero,
since ␭ 0 ( ⑀ ) is a root of the characteristic polynomial. There-
⫹k ⫺2 共 k 4 ⫹k ⫺4 兲 , 共A6兲 fore we are left with
␴ 1 ⫽k 1 关 T 兴共 k 2 k 3 ⫹k 2 k 4 ⫹k ⫺2 k 4 ⫹k 3 k 4 兲 ⫹ 共 k ⫺1 k ⫺2 ⳵␭0 ⳵␴ 0
␴1 ⫹ ⫽0, 共A13兲
⫹k ⫺1 k 3 ⫹k 2 k 3 兲共 k 4 ⫹k ⫺4 兲 , 共A7兲 ⳵⑀ ⳵⑀

␴ 0 ⫽k 1 关 T 兴 k 2 k 3 k 4 共 1⫺e ⫺ikp 兲 . 共A8兲 where it is assumed that all derivatives are evaluated at ⑀
⫽0. From Eq. 共A13兲, we get a⫽⫺( ⳵␴ 0 / ⳵ ⑀ )/ ␴ 1 . The coef-
Letting ⑀ ⫽ikp, the roots of Eq. 共A4兲 are given by ficient b is readily found from the second derivative,
␭ 0 ⫽a ⑀ ⫹be 2 ⫹O 共 ⑀ 3 兲 .
The real parts of the other eigenvalues are all negative at
⑀ ⫽0. The effective spatial diffusion, D, is computed from
共A9兲
b⫽
1 ⳵ 2␭ 0
2 ⳵⑀2
⫽⫺
1 ⳵ 2␴ 0
2␴1 ⳵⑀ 2
⫹2 ␴ 2
⳵␭0
⳵⑀
冋 冉 冊 2

⫹2
⳵␴ 1 ⳵ ␭ 0
⳵⑀ ⳵⑀
. 册
共A14兲
the decay rate of the long wave modes k⬇0 of P⫽ P 1 ⫹ P 2
We find
⫹ P 3 ⫹ P 4 . The eigenvectors corresponding to the eigenval-
ues ␭ 1 , ␭ 2 , ␭ 3 are such that P⬇0, so that the decay rates of k 1关 T 兴 k 2k 3k 4
the corresponding eigenmodes do not affect the magnitude of a⫽⫺ , 共A15兲
␴1
long wave modes of P. Thus, these three eigenvalues are not
related to D. The eigenvalue ␭ 0 is responsible for the evo-
lution of long wave modes of P. When the eigenvector cor-
responding to ␭ 0 is taken as the initial value for
b⫽⫺
a
2 冉␴2
1⫹2 a .
␴1
冊 共A16兲

关 P 1 , P 2 , P 3 , P 4 兴 in Eq. 共A1兲, the solution to the master equa- We recognize in the first equation part of the formula for
tion, Eq. 共6兲, is given by velocity, Eq. 共38兲, so that v ⫽⫺a p. From Eqs. 共A11兲 and

冋 册P 1 共 n,t 兲 共A12兲, and its definition, Eq. 共43兲, we obtain the expression
P 2 共 n,t 兲
P 3 共 n,t 兲
P 4 共 n,t 兲
⫽exp ik np⫹冋冉
I共 ␭ 0 兲
k
t ⫹k 2
k2

R共 ␭ 0 兲
t 册 for randomness,

r⫽1⫹2
␴2
a. 共A17兲

冋 册
␴1
c 1共 0 兲 This formula, together with the expressions for ␴ 1 and
c 2共 0 兲 ␴ 2 , Eqs. 共A6兲, 共A7兲, has been used throughout the paper.
• , 共A10兲
c 3共 0 兲
c 4共 0 兲
F. Jülicher, A. Ajdari, and J. Prost, Rev. Mod. Phys. 69, 1269 共1997兲.
1
where I(␭ 0 ) and R(␭ 0 ) are, respectively, the imaginary and 2
A. Parmeggiani, F. Jülicher, A. Ajdari, and J. Prost, Phys. Rev. E 60, 2127
real parts of the eigenvalue ␭ 0 . The imaginary part is re- 共1999兲.
sponsible for the propagation 共average velocity兲 of long
3
A. B. Kolomeisky and M. E. Fisher, J. Chem. Phys. 115, 7253 共2001兲.
4
G. Lattanzi and A. Maritan, Phys. Rev. E 64, 061905 共2001兲.
wave modes, and the real part is responsible for the decaying 5
D. Keller and C. Bustamante, Biophys. J. 78, 541 共2000兲.
共dissipation兲 of the long wave modes. From Eqs. 共A10兲 and 6
B. Derrida, J. Stat. Phys. 31, 433 共1983兲.
共A9兲, we obtain the average velocity, 7
R. F. Fox and M. H. Choi, Phys. Rev. E 63, 051901 共2001兲.
8
N. G. van Kampen, Stochastic Processes in Physics and Chemistry
V⫽⫺ap, 共A11兲 共Elsevier, Amsterdam, 1981兲.
9
A. Giacometti, A. Maritan, and A. L. Stella, Int. J. Mod. Phys. B 5, 709
and the effective diffusion constant in 共nm2/s兲 共1991兲.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07
J. Chem. Phys., Vol. 117, No. 22, 8 December 2002 Kinetics models for motor proteins 10349

10
J. Howard, Mechanics of Motor Proteins and the Cytoskeleton 共Sinauer 23
M. J. Schnitzer, K. Visscher, and S. M. Block, Nat. Cell Biol. 2, 718
Associates, New York, 2001兲. 共2000兲.
11
B. Hille, Ionic Channels of Excitable Membranes 共Sinauer Associates, 24
C. M. Coppin, D. W. Pierce, L. Hsu, and R. D. Vale, Proc. Natl. Acad. Sci.
New York, 1984兲. U.S.A. 94, 8539 共1997兲.
12
R. D. Astumian and M. Bier, Biophys. J. 70, 637 共1996兲. 25
K. Svoboda, P. M. Mitra, and S. M. Block, Proc. Natl. Acad. Sci. U.S.A.
13
A. Mogilner, T. Elston, H. Y. Wang, and G. Oster, in Joel Keizer’s Com-
91, 11782 共1994兲.
putational Cell Biology, edited by C. P. Fall, E. Marland, J. Tyson, and J. 26
Wagner 共Springer, New York, 2001兲, pp. 321–380. H. Wang, T. Elston, A. Mogilner, and G. Oster, Biophys. J. 74, 1186
14
S. Rice, A. W. Lin, D. Safer et al., Nature 共London兲 402, 778 共1999兲. 共1998兲.
15
G. Lattanzi and A. Maritan, Phys. Rev. Lett. 86, 1134 共2001兲.
27
J. Howard, Annu. Rev. Physiol. 58, 703 共1996兲.
16
R. D. Vale and R. A. Milligan, Science 288, 88 共2000兲. 28
The exponential factor coming from the GDB tends to 1 in the ⌬x→0
17
M. Nishiyama, E. Muto, Y. Inoue, T. Yanagida, and H. Higuchi, Nat. Cell limit.
Biol. 3, 425 共2001兲. 29
Indeed, the renormalization approach presented in the previous section
18
A. Mogilner, A. J. Fisher, and R. J. Baskin, J. Theor. Biol. 211, 143 does not take into account the discontinuities in the potential; a correction
共2001兲. to the model can be obtained by a careful insertion of waiting time distri-
19
Z. Koza, J. Phys. A 32, 7637 共1999兲.
butions, that is the s ⬘ at the discontinuity sites does not have a simple
20
Z. Koza, Physica A 285, 176 共2000兲.
21
K. Visscher, M. J. Schnitzer, and S. M. Block, Nature 共London兲 400, 184 renormalization as the ones along the slope. Nonetheless we will restrict
共1999兲. our analysis only to simple jump processes.
22
M. L. Moyer, S. P. Gilbert, and K. A. Johnson, Biochemistry 37, 800
30
Howard 共Ref. 27兲 suggested that detachment may serve to prevent kinesin
共1998兲. motors from wasting ATP when faced with insurmountable loads.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.236.27.111 On: Mon, 15 Dec 2014 14:41:07

You might also like