Professional Documents
Culture Documents
R. D. A STUMIAN
Abstract Proteins and other macromolecules can act as molecular machines that convert energy from one
or at the single molecule level the probability for a molecule to be in any state “j” fluctuates, and the
probability current from any other state “i” to state “j” is given as the sum of a steady state current and
a “pumped” current, Iij = Iijss + Fij dPj /dt where Fij is the fraction of the fluctuating current into/out of
state j coming directly from state i, and dPj /dt is the rate of change of the probability for the molecule
to be in state “j”. If the fluctuations arise from an equilibrium source microscopic reversibility guarantees
that the time average of the pumped current is zero. If however the fluctuations arise due to the action of a
non-equilibrium source the time average of the pumped current is not in general zero, and can be opposite
in sign to the steady state current. The pumped current provides a mechanism by which fluctuations arising
either from an internal non-equilibrium chemical reaction or from an external driving can do electrical,
mechanical, or chemical work on a system by coupling into the equilibrium conformational transitions of a
protein. In this paper I will review work done on elaborating the mechanism of stochastic pumping and also
discuss a thermodynamically consistent approach for modeling the effects of dynamic disorder on enzymes
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3-Catenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Adiabatic pumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Kinetic mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Developing a thermodynamically consistent model for molecular fluctuations and dynamic disorder 17
Pumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Catalysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
SUMMARY POINTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1 INTRODUCTION
The ability to convert free-energy from one form to another is essential for life. Our cells
store energy in chemical form (often Adenosine Triphosphate, ATP, or Guanosine Triphos-
2
Astumian 3
phate, GTP) and then use the energy released by catalytic breakdown of these molecules
to do various types of work, e.g., to pump ions from low to high electrochemical potential,
lar motors to move material from one place to another (Fig. 1). Despite the fundamental
similarity between these different energy conversion processes, the function of a molecu-
lar motor is typically described in a very different way than is the function of a molecular
Models for molecular motors (39, 64), e.g., have focussed on an ATP driven mechani-
cal power stroke - a viscoelastic relaxation process where the protein starts from a “non-
equilibrium, strained” conformation due to the action of ATP at the active site. The sub-
sequent relaxation following product (ADP or Pi) release can be visualized much as the
In contrast, molecular pumps are most often modeled in terms of chemical kinetics (46),
where ATP energy is used to change the relative affinities of and barrier heights between
binding sites by sequentially favoring different conformational states of the protein as ATP
is bound, hydrolyzed, and the products released. The conformational relaxation and molec-
The perspective I will develop in this review is that at the single molecule level molec-
ular machines are mechanically equilibrated systems that serve as conduits for the flow of
and the environment on which work is to be done (9). Consider the process
kBj
⋯o /Bo /⋯o /Ao kAB
/Bo /⋯o /Ao /⋯
kAi kBA
where A and B are states in the conformational cycle of one of the free-energy transduction
processes shown in Fig. 1. We normally focus on the steady state probability current
ss
IAB = kAB PAss − kBA PBss to assess the average direction of cycling - whether the net flow
is to the left or to the right. In a fluctuating environment or at the single molecule level,
however, we must add a term that reflects the correlations between the fluctuating rates and
4 Stochastic Pumping
fluctuating probabilities
dPB (t)
IAB (t) = IAB
ss
(t) + FAB (t) . (1)
dt
where FAB , is the fraction of the “transient” change in the probability to be in state B,
dPB /dt, coming directly from (to) state A. (5, 8, 38). If the fluctuations in FAB and in
PB are uncorrelated the long time average of the second term, the “pumped” current, is
zero. If, however, the system is pumped, e.g. by simultaneously raising and lowering the
free energy of state B and the energy barrier between B and A (and hence modulating kAB
and kBj , but with the product kAB kBj constant) the average pumped flux is positive since
FAB (t) is greater than 1/2 when PB is increasing and FAB (t) is less than 1/2 when PB is
decreasing (17).
the same irrespective of how different the probability ratio PA /PB is from the equilibrium
constant KAB = kAB /kBA or whether the system is pumped or not. The conformational
transitions in the cycles by which these machines carry out their function are intrinsically
“equilibrium” processes between states that are very close to thermal equilibrium. A time-
relative stabilities of the states in the cycle and the rate constants for the conformational
transitions between them in a correlated way thereby driving net flux through the cycle and
nane based molecular motor (47). Catenanes are molecules with two or more interlocked
rings. Fig. 2a shows a 3-catenane (three interlocked rings) and Fig. 2b illustrates a 2-
catenane (two interlocked rings). The salient feature of these two molecules is that binding
sites or bases (the blue, red, and green boxes labelled 1,2, and 3, respectively) for the
Astumian 5
small purple rings can be designed and located on the large orange ring. The bases can
be designed such that their interaction energies with the purple rings can be independently
sequence of external cyclical changes to the interaction energies of the sites can squeeze
the purple rings to undergo directional rotation by a mechanism similar to peristalsis (6)
but where the transitions occur by thermal noise so the system operates as a Brownian
2.1 3-Catenance
Let us first consider the three ring catenane shown in Fig. 2 a). The larger yellow ring has
three distinct recognition stations, labeled 1,2, and 3, for the two identical purple rings.
The purple rings cannot pass one another, nor can they occupy the same station, as they
make thermally activated transitions from one station to another. Thus there are a total of
three distinguishable states, labeled A,B, and C. The interaction between a purple ring with
breaking the interaction between one station and one ring. For example the transitions
from state A to state B requires breaking the interaction of the ring on site 1, as does the
transition from state C to B. Using this analysis the rate constants for the transitions are
‡
kAB = kCB = Aeβ(E1 −E )
‡
kBC = kAC = Aeβ(E3 −E )
(2)
‡
kCA = kBA = Aeβ(E2 −E )
where A is a frequency factor, β = (kB T )−1 is the inverse of the thermal energy, and
E ‡ > 0 is the energy of the barrier between the stations, which we assume to be the same
slowly that the state probabilities are given by an equilibrium relation (Pi /Pj = kji /kij for
6 Stochastic Pumping
i, j = A, B, C) at every instant, can drive directional rotation of the small rings about the
larger ring. Beginning with E1 , E3 << E2 where the system is almost certainly in state A
the condition E2 , E3 << E1 . During this process, the purple ring originally on base 1 will
have been transferred by an equilibrium process to base 2 to reach state B. Because the
energy E3 remains very low, the ring on base 3 will not have moved so the transfer can
only have occurred in the clockwise direction. The interaction energy E3 is then slowly in-
this process the ring on base 3 will have been transferred to base 1, also in the clockwise
direction, to reach state C. Finally, the interaction energy E2 is slowly increased and si-
process the ring on base 2 will have been transferred to base 3 in the clockwise direction
clockwise cycling of each purple ring through the bases in order 1 → 2 → 3 → 1, with two
cycles through the states necessary to give each ring one clockwise turn.
Directional rotation due to very slow modulation may be surprising given that the instan-
taneous steady state current for any transition, e.g. A → B, is zero (IAB
ss
= kAB PAeq −
kBA PBeq = 0) irrespective of the values of E1 , E2 , and E3 . However, when the external
conditions are caused to change, even very slowly, we must account for the changing state
probabilities by adding the fraction of the total “transient” change in the probability to be in
state B coming directly from (to) state A, FAB dPB /dt, to the steady state current (5, 8, 38)
to get the net instantaneous probability current between states A and B, Eq. (1). Since
ss
IAB (t) = 0 at every instant, the net current for cyclic modulation of the energies averaged
over a period is
ω −1 dPB
IAB = ω ∫ FAB (t) dt = ω ∮ FAB (t)dPB (t), (3)
0 dt
Astumian 7
where ω is the frequency of the modulation. Under adiabatic conditions (very slow modu-
ad
lation) the state probabilities are at equilibrium at every instant and the fraction FAB of the
ing,
kBA kAB PAeq
ad
FAB = = (4)
kBA + kBC kAB PAeq + kCB PCeq
ad
constants in Eq. (1). The adiabatic integrated current per cycle ∮ FAB dPBeq - the area
ad
enclosed in a plot of FAB vs. PBeq parametrized by time - is purely geometric and does
ad
not depend on frequency. Further, since both FAB and PBeq can vary at most between 0
and 1, the limit is one cycle through the states per cycle of modulation, and 1 turn for each
ring per 2 cycles of the modulation (8, 30). The rate constants Eq. (1) obey the simple
relation kAB kBC kCA = kAC kCB kBA irrespective of the values of the energies E1 , E2 , E3 .
This relation is a necessary and sufficient condition for the steady state component of
ss
the current to be zero, IAB = 0 but, as we have seen, in a fluctuating environment this
relation between the rate constants, despite assertions in the literature to the contrary, is
not a sufficient condition to assure that the average net cyclic flux, IAB is zero (19, 61, 80).
(Fig. 2c), where the energy E1 is controlled by the redox potential and the energy E2 is
The geometric effect by which pumping is achieved for a 3-catenane is what Michael
Berry has termed the geometric phase (24). A similar picture emerges for a wide variety of
physical phenomena including dissipationless pumping of electrons (14, 72, 73), a mecha-
nism for bio-molecular ion pumps (5, 70), phase control in oscillating chemical reactions
The pumping mechanism for the 3-catenane based molecular motor illustrates an impor-
tant principle for molecular machines that is very different than the mechanism by which
the desired motion, and without the energy there would be no motion at all. For molecular
machines, however, the parts of the machine are constantly moving about even at thermal
equilibrium. The design of the 3-catenane motor focuses on restricting or preventing the
undesired parts of the thermal motion (4,10). In each transition, one of the small rings acts
as an immobile obstacle, thereby setting the direction of motion of the other ring when,
by thermal noise, the mobile ring moves from an unstable site to a stable site. By restrict-
ing the backward motion, only the desired motion remains - a concept often described as
We can better understand this concept by contrasting the case of the 3-catenane with
modulation of the interaction energies between the yellow ring and the bases. For the
2-catenane in Fig. 2b the rate constants out of any state are the same for clockwise and
counterclockwise transition
‡
kAB = kAC = Aeβ(E1 −E )
‡
kBC = kBA = Aeβ(E3 −E )
(5)
‡
kCA = kCB = Aeβ(E2 −E )
ad
With these rate constants we have FAB = 1/2, a constant, and thus there is no possibility
for directional adiabatic pumping. When, e.g., E1 is increased the probability for transi-
tion out of state A to state C is exactly the same as the probability for transition to state B.
Non-adiabatic pumping (14) at higher frequencies is possible for the 2-catenane (30, 60),
however, since the state probabilities appearing in FAB are no longer given by their equi-
librium values and are frequency dependent. A very general and easily implemented com-
Astumian 9
putational scheme for calculation of the instantaneous state probabilities and currents for
a cyclic system with arbitrarily large amplitude and frequency modulation was given by
Robertson and Astumian (61). The 3-Catenane three state model is perhaps the simplest
example of adiabatic stochastic pumping where only binding site energies are modulated.
The directionality requires the interaction between the two purple rings. If in addition to a
binding site energy a barrier energy can be directly modulated a two-state model for pump-
ing is possible (16,56) where both adiabatic and non-adiabatic components of the pumping
can be analytically evaluated. In the next section we will focus on such a two-state model
for a membrane pump which is also analogous to the Michaelis-Menten mechanism for
enzyme catalysis.
ATP driven pumps are proteins that span a cell or organelle membrane and use energy
from adenosine triphosphate (ATP) hydrolysis to pump ligand (often ions such as Na+ ,
K+ , H+ , or Ca++ ) across the membrane, thus generating and maintaining the ion electro-
chemical gradients essential for life (46). In a simple picture of a membrane pump (figure
1b), the protein structure presents energy barriers (gates) for ligand permeation at the two
entrances, one on either side of the membrane, surrounding an energy well (binding site)
in the middle. Fig. 2a shows an energy diagram for this two-barrier, one-site model of an
ion transporter. The differential barrier height u and well energy are internal parameters
(27) controlled by the conformation of the protein and do not influence the overall trans-
port equilibrium ∆µ = µ1 − µ2 . In the absence of input energy, ions flow from high to
low electrochemical potential. Conformational fluctuations of the pump protein cause the
relative energies of the two gates, as well as the binding energy (well depth) for ligand, to
fluctuate - i.e., both (t) and u(t) depend on time through the protein conformation. Un-
driven fluctuations, however, are not correlated and so cannot drive uphill pumping. The
absence of correlations in the fluctuations of (t) and u(t) is reflected in the principle of
10 Stochastic Pumping
When ATP is bound to the protein, hydrolyzed, and product released in chemically
driven pumping the protein undergoes shape changes in which the relative gate and bind-
ing energies for the ion fluctuate in a correlated way, causing transport of ions across the
external oscillating electric field can also drive structural changes of a pump protein (45)
and can cause non-equilibrium correlated modulation of the relative barrier height u(t)
and the well energy (t), thereby driving uphill pumping. This was shown experimen-
tally by Tsong and colleagues (49, 65, 81, 82) who applied a fluctuating external electric
field to suspensions of red blood cells. The zero average applied fields were able to drive
thermodynamically uphill transport via the ion pump sodium, potassium ATPase (Na,K
ATPase) even under conditions where ATP hydrolysis could not occur. Interpretation of
(19, 76–78, 80) that explains how an external oscillating or fluctuating perturbation can
of ligand from low to high electrochemical potential (19), catalysis of a chemical reac-
tion away from equilibrium (17), or performance of mechanical work on the environment
(11). It seems likely that the same conformational motions are exploited in ATP hydrolysis
driven pumping.
We can understand conformational pumping in terms of the simple kinetic model in Fig.
3a. The instantaneous net current between reservoir 1 and the well is
which, by decomposing the currents Ii (t) = Iiss (t) + δIi (t) and observing that I1ss (t) +
ss
dPEL dδpEL
I1 = I1ss + F1 [ + ] , (6)
dt dt
where we suppress the explicit denotation of the time dependence of the quantities, where
steady state value and a deviation from that value, and where F1 = δI1 /(δI1 + δI2 ) is
the ratio of the fluctuating current between reservoir 1 and the well to the total fluctuating
ss
current into/out of the well. The term F1 dPEL /dt is the adiabatic, and the term F1 dδpEL /dt
is the non-adiabatic contribution to the pump current. The derivation can be generalized
to continuous systems using Gauss’ law for converting between the rate of change of a
density within a volume and the total current into the volume (see Fig. 3 d).
From elementary energetic considerations we have the simple relations between pairs of
rate constants and the internal parameters, and u, and the external parameters µ1 and µ2
k1 k2 k−1 k1
= eβ(µ1 −) ; = eβ(−µ2 ) ; = eβu ; = eβ(∆µ−u) . (7)
k−1 k−2 k2 k−2
where ∆µ = µ1 − µ2 . From Eq. (7) we easily derive another relation between all four rate
k1 k2
= eβ∆µ . (8)
k−1 k−2
The kinetic mechanism for ligand transport shown in Fig. 3a is identical to the Michaelis-
Menten (MM) mechanism for catalysis of the chemical reaction L1 ⇌ L2 where k1 and
k−2 are effective first order rate constants into which the concentrations [L1 ] and [L2 ],
respectively, have been subsumed. In this model µ1 and µ2 are the chemical potentials of
substrate, L1 , and product, L2 , respectively. The relative gate height u has the interpreta-
tion of the chemical specificity (relative lability) of the enzyme - when u < 0 the enzyme
of L2 ) and when u > 0 the enzyme is specific for L2 (i.e., the binding/dissociation of L2 is
12 Stochastic Pumping
faster than the binding/dissociation of L1 ) . The well depth specifies the binding affinity
dPEL
= −τ −1 PEL + (k1 + k−2 ) (9)
dt
where τ = (k1 + k−1 + k2 + k−2 )−1 is the relaxation time for ligand binding in the well.
The instantaneous steady state probability for the well to be occupied is obtained by set-
ting dPEL /dt = 0 in Eq. (9) and solving for PEL to find PEL
ss
= (k1 + k−2 )τ . The
I1ss = (k1 k2 − k−1 k−2 ) τ = (1 − eβ∆µ )k2 k1 τ . If the internal parameters u and fluctuate,
ss
the rate constants and hence, PEL , F1 , and I1ss , all vary in time. Nevertheless, irrespec-
tive of the instantaneous values of u(t) and (t), the sign of I1ss is determined solely by
∆µ = µ2 − µ1 . This is not true however of the pumped current. If we follow the sequence
(u < 0) Ð→ ( < µ1 ) Ð→ (u > 0) Ð→ ( > µ2 ) Ð→ (u < 0), etc. the pumped current will
be positive (from reservoir 1 to reservoir 2) even though µ2 > µ1 . The maximum proba-
bility for an ion to be pumped in one cycle is achieved in the limit ∣u∣ → ∞. In this case
the steady state current is nearly zero because one of the gates is very high at every instant.
The probability to pump an ion then is the difference in occupancy between the state where
( < µ1 ) and that where ( > µ2 ). This probability can be written tanh [β(∆ − ∆µ)/2]
since the well equilibrates with the reservoir to which it has finite access. Thus the max-
imum average output energy per cycle is Emax, out = ∆µ tanh [β(∆ − ∆µ)/2], the min-
imum input energy per cycle is Ein,min = ∆, with the limiting thermodynamic efficiency
(5, 6)
∆µ
ηmax = tanh [β(∆ − ∆µ)/2]. (10)
∆
For an input energy of ∆ = 20β −1 (i.e., the energy provided by ATP hydrolysis under
In order to compare the theory for stochastic pumping with experimental results of Tsong
and colleagues (49) consider a situation in which the internal parameters change periodi-
cally in time with frequency ω. For small amplitude oscillations u(t) = u0 + δu cos (ωt)
and (t) = 0 + δ cos (ωt − φ), Eq. 6 can be evaluated in the small perturbation limit (16)
sin (φ)̃
ω + cos (φ)̃
ω2
I1 − I1ss = K (11)
1+ω ̃ 2
where K is a constant that is proportional to the product δδu. In Fig 3c the data of Liu,
Astumian, and Tsong (49) for ac field induced pumping of both Rb+ (an analogue of K+ )
and Na+ by the Na, K − ATPase as functions of the reduced frequency are fit to Eq. (17),
with τ = 10−6 s for sodium and τ = 10−3 s for rubidium based on the optimal pumping
frequencies observed in the experiments. The fit parameter φ = π/2.1 is very close to π/2
suggesting that the non-adiabatic contribution is essentially negligible except at very high
̃ >> 1. Thus we conclude that the Na,K ATPase may work in many respects
frequencies ω
like an adiabatic pump, where two internal parameters are caused by the applied field to
the oscillating external field ψ(t), can cause two internal parameters to oscillate out of
phase with one another. There are two “major” conformational states, EA and EB . State
well and the reservoir 1 (uA < 0, and F1,A > 1/2) and the other, state EB , has low affinity
(uB > 0, and F1,B < 1/2). If EA and EB have different dipole moments, an external ac
field will alternately favor first one, and then the other state, causing the average values of
ss
F1 and of PEL to oscillate. Let the conformational transition be governed by two relaxation
times, one “fast” relaxation time that govens u and the other “slow” relaxation time that
14 Stochastic Pumping
ss
governs . As a result of the different relaxation times F1 , PEL , and δpEL will oscillate out
ss
of phase with one another. The phase lag between F1 and PEL is caused by an internal
Even at low frequency the system is not in global equilibrium, but only in equilibrium with
which δpEL fluctuates out of phase with F1 has also been discussed (16, 18).
In the experiment of Tsong et al. (49) the conformational oscillation was driven by an
applied oscillating electric field. In chemically driven pumping, where e.g. ATP hydrolysis
drives transport, the stochastic binding of reactants and release of products cause transitions
between states of the protein. In this case, after phosphorylation or dephosphorylation, the
differential barrier height that controls the parameter F1 rapidly approaches its final value,
ss
followed by a slower relaxation of the well energy (i.e., PEL ) to its new value. In this way,
a stochastic input (ATP hydrolysis) is converted into two on average phase shifted outputs.
Such hysteretic behavior is very general in proteins or for that matter for any relatively
A simple model (18) for stochastic pumping by a single external parameter ψ(t) is shown
in Fig. 4a alongside a kinetic model for ATP driven pumping involving the same protein
states in Fig. 4b. The model in Fig. 4a can also be written in the form
L1 +O EA o
k1A
/E Lo k2A
/ E +L (12)
k−1A O
A
k−2A
A O 2
L1 + EB o
k1B
/E Lo k2B
/ E +L
B B 2
k−1B k−2B
The sets of rate constants kiA and kiB (i = ±1, ±2) separately satisfy Eqs. (7) and (8),
with uA and A , and with uB and B , respectively. Additionally there is a detailed balance
Astumian 15
condition (77)
k2B k−2A αAB γBA k−1B k1A
= = (13)
k−2B k2A αBA γAB k1B k−1A
that constrains how the rate constants for fluctuation between the two states can be as-
signed. Considering the case in Fig. 4, neither EA nor EB are very good catalysts - there
is one large barrier preventing free transport of ligand for each form. The possibility of
fluctuations between the two forms can have a significant beneficial effect on the catal-
ysis. If the α’s and γ’s are large (meaning fast fluctuations), the transporter is a much
better catalyst than either of its two conformational states alone! By using the pathway
+L1 −L2
⇌ ⋯EA ⇌ EA L ⇌ EB L ⇌ EB ⇌ EA ⋯ ⇌ ligand can pass between reservoir 1 and 2
without surmounting a large activation barrier. The flow of ligand however is from high
to low chemical potential - the protein is just a catalyst, but a better catalyst because of
γAB = γAB
0
e−βψ(t)/2 , γBA = γBA
0
eβψ(t)/2 , where ψ(t) is any autonomous function of time
whatsoever (19), there is net flow of ligand from reservoir 1 to reservoir 2 when ∆µ = 0.
Note that ψ drops out of the product αAB γBA /(αBA γAB ) so the detailed balance condi-
The protein is a poor catalyst in state EA and a poor catalyst in state EB . Equilibrium
conformational fluctuation between the states turns the protein into a better catalyst. These
same conformational transitions in the presence of external driving or involved in the catal-
ysis of a non-equilibrium chemical reaction such as ATP hydrolysis allow the protein to
function as a free-energy transducer that harvests energy from the external driving or ATP
Sunney Xie and colleagues have recently provided compelling experimental evidence at
the single molecule level that conformational fluctuations are important for enzymes (50).
This is consistent with much work on the general importance of conformational flexibility
in enzyme catalysis reviewed in (36) and with a general model for enzyme conformational
16 Stochastic Pumping
flexibility given in (23) and role of dynamics in protein function (22). How can we theoret-
ically model the effects of internal fluctuations, present even at equilibrium, on enzymes?
formed when a long rod shaped molecule is threaded through a macrocylic ring compound
and then “stopper” groups added to the ends of the rod to prevent the macrocycle from
escaping (see Fig. 5). As with the catenane discussed in Fig. 2, separate binding sites for
the macrocycle can be chemically incorporated on the rod. We consider two such sites in
the model in Fig. 5, where there is a steric barrier between the two sites hindering but not
preventing exchange of the ring between sites 1 and 2. At equilibrium the occupancy of the
macrocycle at the two sites is determined by the relative interaction energies of the sites.
There are two ways in which the relative occupancy can be shifted away from the equi-
librium value. These are illustrated here by a “Maxwell’s demon”, an “intelligent being”
that uses information about the location of the macrocycle to determine when to open and
close a gate, and a “Smoluchowski’s demon” drawn as a blindfolded being that randomly
raises and lowers the interaction energy between the macrocycle and one of the binding
sites (29).
In the Maxwell’s demon case, if the demon, spotting the location of the ring, lowers
the barrier when the ring is on station 1, and raises the barrier when ring is on station 2,
the ring will obviously spend more time on station 2 than on station 1 despite the fact that
the two station have identical interaction energies at every instant. Just this scenario has
activated barrier (1), where the sensitivity of the trigger mechanism depended on the loca-
tion of the macrocyclic ring. The raising and lowering of the gate illustrates one of the key
design principles of a brownian motor or stochastic pump of bilability (9) which has been
Astumian 17
mon who randomly raises and lowers the interaction energy for one of the sites, thereby
switching the relative stabilities of sites 1 and 2 back and forth. When site 1 is less sta-
ble escape to site 2 is rapid, whereas when site 2 is less stable , escape to site 1 is less
rapid. As a result of the speed of escape the average occupancy of site 2 is greater than the
average occupancy of site 1 under the fluctuating conditions even though on average, the
interaction energies of sites 1 and 2 are the same. Leigh et al. have used this principle of
bistability (coupled with bilability) to design a catenane based rotary molecular motor (?).
The two principles of bistability and bilability are the cornerstones of a minimal brown-
ian motor or stochastic pump (9). Roughly speaking, the lability of a kinetic pathway is
reflected in the term Fij in Eq. (1), and the switching of the stability is reflected in the
term dPj /dt in Eq. (1). The fact that a single term ψ(t) can drive a system away from
equilibrium even if ψ fluctuates randomly forces us to ask how we can describe, in a single
theory, both the assuredly present equilibrium fluctuation in ψ and also non-equilibrium
driving of ψ.
Consider the Michaelis-Menten scheme with rate constants that depend on some “con-
k1 eβaψ(t) k2 e−βaψ(t)
L1 + E o / EL o / E + L2
k−1 eβbψ(t) k−2 e−βbψ(t)
Note that in the ratio Eq. (8), ψ cancels in the numerator and denominator. When a = b the
system is set up for a “Maxwell’s demon” who, by increasing ψ when PEL is greater than
average, and decreasing ψ when PEL is less than average drives pumped current from L1 to
L2 even when ∆µ < 0. When a ≠ 0 and b = 0 the system is set up for a “Smoluchowski’s
demon” who by causing random fluctuations also drives pumped current from L1 to L2
18 Stochastic Pumping
(16–19). How can we model both equilibrium and non-equilibrium fluctuations in ψ(t)?
The equation for the rate at which the bound state probability changes can be written
It is tempting to consider the model proposed by Zwanzig (84) and subsequently adopted
by Wang and Wolynes (79), Schenter, Lu, and Xie (63), and Lerch, Rigler, and Mikhailov
(48), among others, to describe the effects of “dynamic disorder” on proteins, including
control variable” (e.g., ψ(t)) that influences the dynamics of the rate process dPEL /dt
but which is not influenced by the value of PEL , so that the equation of motion for ψ(t) is
given by the autonomous Langevin equation dψ/dt = −λ ψ+ξ(t), where ξ(t) is taken to be
white noise. This picture, however, is not thermodynamically consistent for endogenous
(internal) noise as had been previously pointed out by Astumian et al. (19). The joint
trajectories [PEL (t), ψ(t)] do not obey microscopic reversibility and when ∆µ = 0 there
is still net pumped flux from reservoir 1 to reservoir 2 in violation of the second law of
reaction of the enzyme state on the likelihood for the control parameter to adopt some
particular value (19) by augmenting the equation of motion for ψ(t) to read (11, 52)
dψ
= g(PEL , ψ) + fSD (t) + fMD (PEL ) + ξ(t) (15)
dt
where we require
The autonomous function fSD (t) describes the action of “Smoluchowski’s demon”, and
the function fMD (PEL ) describes the action of “Maxwell’s demon”. When fSD (t) =
fMD (PEL ) = 0 (when the demons are at rest) the trajectories [PEL (t), ψ(t)] obey mi-
croscopic reversibility, the direction of the enzyme reaction is given solely by the chemical
potential difference ∆µ, and the net flux is zero when ∆µ = 0. The choice of g(PEL , ψ)
Astumian 19
defined in Eq. (16) assures that the curl of the vector field of the two reciprocally coupled
With a time dependent forcing fSD (t), or when there is a mechanism by which ψ(t)
changes depending on whether the active site is or is not occupied fMD (PEL ) , the in-
teraction between ψ(t) and the protein conformational transitions on which it acts allow
free-energy to be transduced from the source of the fluctuation fSD (t) or fMD (PEL ) to do
TRANSITIONS
Microscopic reversibility and its corollary, detailed balance, are among the most important
molecule level, and yet there is great confusion in the literature concerning these principles
and their applicability to non-equilibrium systems (26). In this section I will examine
how microscopic reversibility constrains possible designs for molecular machines, first by
Myoglobin is one of the most well studied proteins (20, 58), and one of the first for which
the x-ray crystal structure was determined (44). When oxygen or carbon monoxide bind
which the iron atom is out of the plane of the heme to a configuration in which the iron
Thinking about the reverse of the binding process, it is tempting to imagine a scenario
in which oxygen dissociates followed by return of the heme group to its original out-of-
the initial equilibrium conformation for the protein where the binding site is unoccupied.
Indeed, this picture seems to be well supported by experiments on ligand dissociation from
and the subsequent conformational relaxation is studied. These investigations reveal that
after light induced dissociation the myoglobin molecule undergoes a local rearrangement
followed by a global conformational change in what has been termed a “protein quake” (2).
The mechanism for photolytic dissociation is shown in Fig. 5 b). The photolytic mech-
anism is clearly not the microscopic reverse of the reaction by which binding of oxygen
reversibility we must have for the most probable pathway the microscopic reverse of the
binding reaction, i.e., for thermally activated dissociation we have the mechanism shown
in Fig. 5 c). These two different mechanisms for dissociation of ligand from the heme
in the International Union of Pure and Applied Chemists (IUPAC) Compendium of Chem-
exactly the reverse of the mechanism in the other direction. This does not apply to reac-
The idea that, following either binding or dissociation of ligand at the heme, the conforma-
tional rearrangement of the protein starts locally and propagates through the protein until
the global change to the new equilibrium conformational state has occurred, follows very
naturally from macroscopic analogy. When we insert our finger into water, ripples prop-
agate from our finger outward. When we remove our finger ripples once again propagate
outward from where our finger had been. This picture was explicitly suggested by Ansari
Astumian 21
et al. (2) for ligand association/dissociation to myoglobin. These authors stated that “bind-
ing or dissociation of a ligand at the heme iron causes a protein-quake”, with the heme as
the focus of the quake. Such a picture is not consistent with microscopic reversibility and
binding causes a quake propagating outward from the focus, then, counter-intuitive though
it may be, thermally activated dissociation must arise by an inward propagating “un-quake”
Conformational transitions - shape changes - of a protein (or any other deformable body
(59, 67)) cause the center of mass of the protein to move relative to the fluid in which the
reversibility such that the backward reaction (e.g., mechanism Fig. 5c)) is the microscopic
reverse of the forward reaction (e.g., Fig. 5 a)) give rise to a reciprocal process (“cycle”)
(Fig. 5d) that according to Purcell’s scallop theorem (59) cannot, in the absence of inertia,
cause net directed motion in a cycle of the forward and backward transitions. Whatever
is done in the forward process is undone in the backward process. Thus, the thermally
On the other hand, although there are doubtless many practical reasons that it is not
biologically relevant for myoglobin, there is no fundamental reason that a cycle of ther-
mally activated binding and photochemically induced dissocation of a ligand could not
provide an effective mechanism for propulsion under the right circumstances. The combi-
nation of Figs. 5a and 5b is shown in Fig. 5e where it is apparent that the conformational
relaxation following photo-assisted dissociation is not the microscopic reverse of the con-
solution can, and in general will, lead to directed motion (67), whether it be of an ion
across a membrane, stepping along a polymeric track, or self-propulsion through the aque-
22 Stochastic Pumping
ous solvent (62). A particularly easy to understand example called the pushmepullyou has
be proposed and discussed by Avron (21). Let us now consider how such non-reciprocal
In a recent paper Togashi and Mikhailov (75) proposed that a polymer, described as an
binding. The binding was modeled by forming elastic links between the ligand and nearby
nodes of the elastic network and allowing the network to relax to its new conformational
energy minimum. The ligand was then removed (the elastic interactions were deleted)
and the system again allowed to undergo conformational relaxation. The overall process
resulting from adding ligand, relaxation, removing ligand, relaxation, adding ligand, etc.
was described by a simple cycle shown in Fig. (6a). Thermal noise was not included in
over-damped elastic relaxation processes. The mechanism is robust. Trajectories begun off
not consistent with microscopic reversibility if the ligand that binds to state EA is the same
molecule as the ligand that dissociates from state EB L . Clearly, the energy of stateEA is
less than that of state EB , and the energy of state EB L is less than that of state EA L. No
matter what fixed arrangement of the energies of the bound states relative to the energies
of the non-bound states the overall cycle of binding ligand to EA , relaxing to EB L, and
releasing ligand from EB and relaxing back to EA needs energy which cannot be provided
by the binding and release of the same ligand under the same conditions.
In Togashi and Mikhailov’s work, the ligand dissociating from the polymer was im-
plicitly different than the ligand that had associated. The details of the ligand binding
EA → EA L and dissociation EB L → EB were not explicitly discussed, nor was the role
Astumian 23
mechanical cycle (40) so long as the motion on each of the two potentials is downhill,
aW < ∆Ufree and (1 − a)W < ∆Ubound , where ∆Ufree = UEB − UEA and ∆Ubound =
UEA L − UEB L .
Our goal is to understand how molecular motors convert chemical energy into non-reciprocal
conformational cycling, and hence into directed motion and mechanical work, in solution
at room temperature where thermal noise is very strong and there is a continual, reversible
exchange of energy between each polymer molecule and its environment. This fact has
extremely important ramifications for how we should describe and think about molecular
motors.
When we look at the mechanism in Fig. 6a it is tempting to term the elastic relaxation
we can compare the power dissipated during the “power stroke” with the power that is con-
tinually and reversibly exchanged between the polymer and the environment to gauge the
relative importance of mechanical vs. thermal effects. If at some point on its energy profile
the polymer experiences a very large force of 100 pN that at that instant moves the center of
mass of the polymer with the very large velocity of 1 m/s, the power instantaneously being
dissipated by the “power stroke” is 10−10 J/s. In contrast, the power reversibly exchanged
time ≈ 10−12 s or 4 × 10−9 J/s, 40 times greater than the maximum power dissipated dur-
ing even a very powerful molecular power stroke! (9). Further, as recognized by Andrew
Huxley (41), the “un-power strokes” EB L → EA L and EA → EB will also occur with ap-
preciable rates in a thermal environment. This was recently shown experimentally (57)
24 Stochastic Pumping
for a simple rotaxane molecule. The ratio of the probability for an uphill fluctuation to
a downhill relaxation by the microscopic reverse process is given by the simple relation
(7, 25)
P (EB → EA ) P (EA L → EB L)
= eβ(∆Ufree −aW ) ; = eβ(∆Ubound −(1−a)W ) . (17)
P (EA → EB ) P (EB L → EA L)
changes by which chemically driven molecular motors move are equilibrium processes.
The only difference between “equilibrium” and “non-equilibrium” for a thermally activated
mechanism is that away from equilibrium the probability to bind ligand when the polymer
is in state EA is different than the probability to bind ligand when the molecule is in state
EB . The physical motions of the molecule that follow binding of ligand are exactly the
The fact that the energy difference between the bound states and the non-bound states
depends on the chemical potential of ligand immediately suggests two approaches for how
to use ligand binding and dissociation to drive non-reciprocal cycling of the polymer con-
5.4 Pumping
tions or fluctuations of the concentration of the ligand between a large and small concen-
tration. The large concentration [L]high > Kd , favors binding of ligand to EA followed by
elastic relaxation EA L → EB L, i.e., binding by the induced fit pathway. The small concen-
tration [L]low < Kd favors dissociation of ligand from EB L followed by elastic relaxation
constant Kd is that concentration at which half the polymer is bound and half is free.
the ligand concentration, although in principle, if the ligand were some intermediate in an
oscillating chemical reaction such as the Belousov-Zhabotinsky reaction (33) the process
The pumped energy flux into the system that allows work to be done on the environment
comes from the fact that, on average, ligand is bound while the chemical potential is high,
and dissociates when the chemical potential is low. Through a cycle of oscillation of the
ligand concentration the energy available to drive flux through the conformational cycle is,
at most, equal the amplitude of the oscillation of the chemical potential, but for very large
oscillation between very low ([L] << Kd ) and very high ([L] >> Kd ) levels the ratio of the
P (EA → EA L → EB L → EB → EA )
= eβ(∆Ufree +∆Ubound −W ) . (18)
P (EA → EB → EB L → EA L → EA )
[L] >> Kd the polymer relaxes from state EA L to state EB L because ∆Ubound > 0, and
as soon as L dissociates when [L] << Kd the polymer relaxes from state EB to state EA
5.5 Catalysis
The second approach for driving directional cycling is catalysis. If a polymer can be de-
are not equal, µL1 ≠ µL2 , the polymer will most likely bind whichever of L1 and L2 has
the higher chemical potential and release whichever of L1 and L2 has the lower chemical
potential. Thus, catalysis autonomously achieves the “bind high release low” by which
pumping drives directional cycling. Note that, since L1 and L2 must be related chemically
(i.e., they are interconvertible), either L1 or L2 can bind to either state EA or state EB , but
with possibly different rates. We can express the ratio of the probabilities for clockwise
26 Stochastic Pumping
and counterclockwise conformational cycling in terms of only the ratios of “off” rate con-
stants for L1 and L2 from EA and EB , the chemical potential difference ∆µ = µL1 − µL2 ,
and the work, W, on the environment required for non-reciprocal conformational cycling
where si = koff,i
L1
/koff,i
L2
, i = A, B. In contrast to the case of external pumping, for catalyt-
ically driven conformational cycling this ratio is independent of the elastic energy differ-
ences ∆Ufree and ∆Ubound . We can think of the mechanism as a chemically driven “in-
formation ratchet” (1, 13) where the active site acts as “Maxwell’s demon” (29), selecting
for L1 in state EA , and for L2 in state EB . Note however that irrespective of how strongly
asymmetric the selectivity of the active site is, the mechanism fails to drive directed trans-
well be the mechanism for enhanced diffusion during active catalysis by an enzyme (55),
where the rapid rotational rearrangement prevents the appearance of net directed motion.
6 CONCLUSIONS
The transitions within the conformational cycles by which molecular machines function are
equilibrium processes. The physical motions of the individual molecules are exactly the
same at and away from statistical equilibrium. Conformational pumping occurs when an
external source or an energy releasing chemical reaction entrains these equilibrium motions
and allowing for a non-reciprocal conformational cycle by which directed motion occurs.
Unsatisfying though it may be, the best description of the mechanism by which, e.g., ATP
hydrolysis drives a molecular motor to move in one direction along a biopolymeric track
is mass action. The fact that binding ATP, converting it to ADP and Pi at the active site,
and releasing ADP and Pi is more likely than the reverse when the ATP hydrolysis reaction
Astumian 27
the protein giving rise to correlations that drive directed motion, pumping ligand across
does not involve judo throws, nor does it bear any resemblance whatsoever to a toy steam
engine. Perhaps the best macroscopic analogy is with an Archimedes screw, a device for
pumping water. In a perfectly constructed Archimedean screw the amount of water pumped
per cycle of turning of the screw is a constant. This geometrical type mechanism can work
effectively only in the very slow limit. However, for molecules, the relevant relaxation
times for the conformational transitions are often a few microseconds. Consequently a
molecular machine that carries out its function several thousand times a second can still be
well within the adiabatic limit. Further, because in a kinetic cycle all states can equilibrate
with one another even if one transition is kinetically blocked, a molecular machine can
do work against significant loads by designing mechanisms that sequentially move kinetic
blockades through the cycle, thereby preventing slip even under the influence of load (8).
Great progress has been made in the design and synthesis of artificial molecular ma-
chines (34,43,54,71) and DNA (35) and small molecule (32) walkers. It is becoming clear
that these are first and foremost molecules, governed by the laws of chemistry rather than
completely specified by the equilibrium energies of the states and by the heights of the
barriers between the states. Chemists have long experience with approaches for control-
ling stabilities and labilities of molecules, and for designing systems with sterically or
energetically hindered pathways allowing for kinetic rather than thermo-dynamic control
of mechanisms. This experience will be crucial in the next steps of interfacing synthetic
7 SUMMARY POINTS
1. Irrespective of how far a system is from statistical equilibrium the instantaneous probabil-
ity flux between any two conformational states of a macromolecule is the sum of a steady
dPj
state current and a pumped current, Iij = Iijss + Fij dt
. In the absence of driving the time
the presence of non-equilibrium driving, however, the average of the pumped current is
not in general zero, and can be opposite in sign and larger in magnitude than the steady
state current.
which an external modulation can do work on the environment. The same conformational
inclusion of a reciprocal reaction between the source of the fluctuation and the protein. A
key test for any model is that in the absence of external driving the joint fluctuations of
4. Far from being predominately mechanical devices slightly perturbed by thermal noise,
molecular motors are overwhelmingly dominated by thermal effects and are first and fore-
most molecules that operate based on the laws of chemistry rather than of macroscopic
mechanics. A key design principle is that molecular machines take advantage of om-
nipresent thermal noise and function by using input energy in part to prevent backward
ACKNOWLEDGMENTS
I am grateful to the German Humboldt foundation for facilitating this work through
Figure Legends
Figure 1: Schematic illustrations of molecular machines that use energy from ATP
hydrolysis to accomplish specific tasks. a) A Molecular pump that move some ligand
enzyme that synthesizes some necessary substance. c) A molecular motor that walks
Figure 2: a) A three ring catenane that can operate as a molecular machine that moves
tions). b) A two ring catenane that can undergo a precise cycle of states in response
ad
to an external stimulus, but where the motion is not directional. c) Plot of FAB = (1 +
eβ(E3 −E2 ) )−1 and PBeq = [1 + eβ(E1 −E2 ) + eβ(E1 −E3 ) ] parametrized by time, with E1 =
−1
the equilibrium probability for state A vs. the fraction of the flux into/out of A that comes
from/to B. The red curve is based on the rate constants for the 2-catenane for which Fad
AB
is constant, and the green curve is based on the rate constants for the 3-catenance for
eq
which Fad
AB is controlled by pH and PA is controlled by the redox potential, where for
Figure 3: a) Potential energy diagram for a membrane pump. The two external pa-
while the two internal parameters, u and fluctuate in time due to conformational fluc-
tuations of the protein. If the fluctuations are driven by, e.g., an oscillating field, the
correlated fluctuations of u and can lead to uphill pumping of ligand. Below the mem-
brane is a kinetic mechanism for the pumping that is analogous to the Michaelis-Menten
mechanism for enzyme catalysis. b) Schematic diagram showing how a single external
30 Stochastic Pumping
parameter, an oscillating field, can cause the internal parameters u and to oscillate out
of phase with one another. The inset shows how the area enclosed by the parametric
ß
plot of F1 vs. PEL is maximized when the external frequency ω matches the system
characteristic frequency τ −1 . c) Fit of data from (49) to the sum of the adiabatic and
non-adiabatic currents Eq. (17). The maximum current was normalized to unity, and the
characteristic frequency for Rb+ pumping (pink squares) was 103 Hz and the character-
istic frequency for Na+ pumping (purple triangles) was taken to be 106 Hz, both cor-
responding to the optimal frequencies for pumping in the experiment. The inset shows
the parametric plots for the adiabatic ( counterclockwise loop on the right) and the non-
adiabatic (clockwise loop on the left) pumping. d) Illustration and general derivation
of the fundamental equation of stochastic pumping based on Gauss’ law. The pumped
current can be written as the product of the fraction of the total fluctuating current that
moves through the specific “channel” of interest (red arrow) multiplied by the volume
integral of the rate of change of the density within the volume. Pumping by using ex-
ternal driving to impose correlations on the fluctuations of the two factors can cause the
sign of the total current to be opposite that of the steady-state current and thereby do
Figure 4: Kinetic diagram showing how an external fluctuating field can entrain the
equilibrium conformational fluctuations of the pump protein to bind ligand from reser-
mechanism showing how, at the single molecule level, ATP hydrolysis can accomplish
the same pumping that the fluctuating electric field causes by alternately phosphorylating
demon for controlling a simple two-state rotaxane based switch. Maxwell’s demon uses
Astumian 31
information about the position of the ring to raise a barrier when the ring is on base
2, and lower the barrier when the ring is on base 1. Even though the interaction energy
between the ring and the two bases is identical at every instant in time, the demon impose
correlation between the height of the barrier and the position of the ring causes the
ring to spend most of the time on base 2. Smoluchowski’s energy demon, which is
blindfolded, raises and lowers the interaction energy between the ring and base 1 at
random, with equal likelihood to raise the energy when the ring is on base 1 and on base
2. At equilibrium of course it would be more likely for the interaction energy of base
1 to fluctuate to a high level (red line) when the ring is on base 2 than when it is on
causing the ring to spend more than half the time on base 2. When the demons are at
rest, i.e., when there is no pumping of the system by an external source, the value of
ψ still fluctuates but in a way that is consistent with microscopic reversibility. Simple
kinetic diagrams illustrating the effects of pumping are shown for the three cases of a
Maxwell’s demon, the demons at rest (no pumping), and Smoluchowski’s demon.
and dissociation gives a reciprocal cycle d) in which the forward and backward processes
are the microscopic reverses of each other and which cannot drive directed motion. On
the other hand, combination of thermally activated binding with photochemically acti-
vated dissociation gives rise to a non-reciprocal cycle e) which can in principle provide
a mechanism for directed motion. f) Energy level diagram for the states involved in the
kB T (fifty times the thermal energy) per transition. These types of conformational tran-
sitions are what Frauenfelder and colleagues (2) call “functionally important motions”,
but they are relevant only for photochemical processes. For thermally activated transi-
tions the energy changes are much more modest (at most around 20 kB T ) and occur at
and allowing viscous relaxation on the “Bound” energy surface, followed by removal
of ligand and relaxation on the “Free” energy surface. Whenever ligand is added or re-
moved, energy is deposited in the system and dissipated during the relaxation. If the
system is set up to harness this energy by attaching the relaxing network to a load, some
of the energy deposited on addition and removal of ligand can by harnessed to do work
(W) in the environment. b) In a thermal environment both forward and backward tran-
sitions are possible. The ratio of the probability for an uphill fluctuation to a downhill
dissipation is a state function (25), Eq. (25). c) If the polymer can bind some substrate
L1 and catalytically convert it to a different molecule L2 , then the energy in the chemical
potential difference of L1 and L2 can drive non-reciprocal cycling of the polymer and
reciprocal cycling by oscillation of the ligand concentration between a low level, where
the polymer is most likely “free”, and a high level, where the polymer is most likely
than the dissociation constant, and the concentration of L2 is less than the dissociation
release L2 , undergo conformational relaxation, bind L1 , etc. For both the pumping and
the catalysis mechanism only unidirectional arrows have been shown. In a thermal envi-
ronment, however, the molecule will occasionally carry out a cycle in reverse.
Astumian 33
L ITERATURE C ITED
1. Alvarez-Perez M, Goldup SM, Leigh DA, Slawin AMZ 2008, A chemically driven
2. Ansari A, Berendzen J, Bowne SF, Frauenfelder H, Iben IET, Sauke TB, Shyam-
sunder E,Young RD 1985, Protein states and proteinquakes. Proc. Natl. Acad. Sci.
276:917–922.
4. Astumian RD 2001, Making molecules into motors. Sci. Am. 285: 45-51.
5. Astumian RD 2003, Adiabatic pumping mechanism for ion motive ATPases. Phys.
to swim in molasses and walk in a hurricane. Phys. Chem. Chem. Phys.. 9: 5067-
5083.
11. Astumian RD, Bier M 1994, Fluctuation driven ratchets: Molecular motors. Phys.
12. Astumian RD, Bier M 1996, Mechanochemical coupling of the motion of molecu-
13. Astumian RD, Derenyi I 1998, Fluctuation driven transport and models of molec-
14. Astumian RD, Derenyi I 2001, Towards a chemically driven molecular electron
15. Astumian RD, Hanggi P 2002, Brownian Motors, Physics Today, 55:(11), 33-39.
16. Astumian RD, Robertson B 1989, Nonlinear effect of an oscillating electric field
17. Astumian RD, Robertson B 1993, Imposed oscillations of kinetic barriers can
cause an enzyme to drive a chemical reaction away from equilibrium. J. Am. Chem.
18. Astumian RD, Chock PB, Tsong TY; Westerhoff HV 1989, Effects of oscilla-
tions and energy driven fluctuations on the dynamics of enzyme catalysis and free-
19. Astumian RD, Tsong TY, Chock PB, Chen YD, Westerhoff HV 1987, Can free-
energy be transduced from electric noise. Proc. Natl. Acad. Sci. USA 84: 434-438.
20. Austin RH, Beeson KW, Eisenstein L, Frauenfelder H, Gunsalus IC 1975, Dynam-
22. Bahar I, Lezon TR, Yan LW, Eran E 2010, Global Dynamics of Proteins: Bridging
24. Berry MV 1990, Anticipations of the geometric phase. Physics Today, 43:(12),
34-40.
Astumian 35
26. Blackmond DG 2009, “If Pigs Could Fly” Chemistry: A Tutorial on the Principle
27. Burbaum JJ, Raines RT, Albery WJ, Knowles JK 1989, Evolutionary optimization
28. Carter NJ, Cross RA 2005, Mechanics of the kinesin step. Nature, 435: 308-312.
29. Chatterjee MN, Kay ER, Leigh DA 2006, J. Am. Chem. Soc., 128, 4058-4073.
30. Chernyak VY, Sinitsyn NA 2009, Robust quantization of a molecular motor motion
E, 77: 050901.
32. Delius M, Geertsema EM, Leigh DA 2010, A synthetic small molecule that can
33. Epstein IR, Showalter K 1996, Nonlinear chemical dynamics: Oscillations, pat-
34. Feringa BL 2007, The Art of Building Small: From Molecular Switches to Molec-
mechanism for autonomous molecular motion. Phys. Rev. Letts., 101: 238104.
37. Hernandez JV, Kay ER, Leigh DA 2004, A reversible synthetic rotary molecular
38. Horowitz JM, Jarzynski C 2009, Exact formula for currents in strongly pumped
39. Howard J 2001, Mechanics of Motor Proteins and the Cytoskeleton. (Sinauer, Sun-
derland, MA).
40. Hugel T, Holland NB, Cattani A, Moroder L, Seitz M, Gaub HE 2002, Single-
41. Huxley AF 1957, Muscle structure and theories of contraction. Prog. Biophys. Bio-
42. Kagan ML, Kepler TB, Epstein IR 1991, Geometric phase shifts in chemical oscil-
43. Kay ER, Leigh DA, Zerbetto F 2007, Synthetic molecular motors and mechanical
44. Kendrew JC, Dickerson RE, Strandberg BE, Hart RG, Davies DR, Phillips DC,
45. Kim YC, Furchtgott LA, Hummer G 2009, Biological Proton Pumping in an Os-
47. Leigh DA, Wong JKY, Dehez F, Zerbetto F 2003, Unidirectional Rotation in a
48. Lerch HP, Rigler R, Mikhailov AS 2005, Functional conformational motions in the
turnover cycle of cholesterol oxidase. Proc. Natl. Acad. Sci. U.S.A. 102: 10807-
10812.
49. Liu DS, Astumian RD, Tsong TY 1990, Activation of N a+ and K + pumping mode
50. Lu HP, Xun L, Xie XS 1998, Single Molecule Enzymatic Dynamics. Science, 282:
1877-1881.
51. Maes C, Netocny K, Thomas SR 2010, General no-go condition for stochastic
52. Magnasco MO 1994, Molecular combustion motors. Phys. Rev. Lett.. 72: 2656-
2659.
53. Markin VS, Tsong TY, Astumian RD, Robertson B 1990, Energy transduction
93: 5062-5066.
54. Michl J, Sykes ECH 2009, Molecular Motors and Rotors: Recent Advances and
55. Muddana HS, Sengupta S, Mallouk TE, Sen A , Butler PJ 2010, Substrate Catalysis
56. Ohkubo J 2008, Current and fluctuation in a two-state stochastic system under
57. Panman MR , Bodis P, Shaw DJ, Bakker BH, Newton AC, Kay ER, Brouwer AM,
1255-1258.
58. Parak FG, Nienhaus GU 2002, Myoglobin, a paradigm in the study of protein
59. Purcell E 1977, Life at low Reynolds number. Am. J. Phys., 45: 3-11.
63. Schenter GK, Lu HP, Xie XS 1999, Statistical Analysis and Theoretical Models of
66. Serreli V, Lee CF, Kay ER, Leigh DA 2007, A Molecular Information Ratchet.
Nature,445, 523-527.
67. Shapere A, Wilczek F 1987, Self-Propulsion at Low Reynolds Number. Phys. Rev.
68. Share AI, Parimal K, Flood, AH 2010, Bilability is defined when one electron is
69. Sinitsyn NA 2009, The stochastic pump effect and geometric phases in dissipative
70. Sinitsyn NA, Nemenman I 2007, The Berry phase and the pump flux in stochastic
71. Siwy ZS, Powell, MR Kalman E, Astumian RD 2006 Eisenberg RS, Negative In-
6: 473-477.
73. Thouless DJ 1983, Quantization of particle transport. Phys. Rev. B 27: 6083.
Astumian 39
74. Tolman RC 1938, The principles of statistical mechanics. Clarendon Press, Reis-
and design principles of molecular machines. Proc. Natl. Acad. Sci., 104: 8697 -
8702.
76. Tsong TY, Astumian RD 1986, Absorption and conversion of electric field energy
bound ATPase transduces energy from dynamic electric fields. Ann. Rev. Physiol.,
50, 273-290.
80. Westerhoff HV, Tsong TY, Chock PB, Chen YD, Astumian RD 1986, How en-
zymes can capture and transmit free-energy from an oscillating field. Proc. Natl.
81. Xie TD, Marszalek P, Chen YD, Tsong TY 1994, Recognition and processing of
82. Xie TD, Chen YD, Marszalek P, Tsong TY 1997, Fluctuation-driven directional
83. Xu D, Phillips JC, and Schulten K 1996, Protein response to external electric fields:
84. Zwanzig RJ 1990, Rate Processes dynamic disorder, Acc. Chem. Res., 23: 148.
40 Stochastic Pumping
ATP ADP+ Pi
L1 + + L2
conform ational cycle
S + P +
conform ational cycle
ATP ADP+ Pi
- +
conform ational cycle
- +
i i+1
Figure 1
Astumian 41
a) A b) A
2 E1,E3 << E2 2 E2,E3 >> E1
1 3 1 3
2 2
2
22
1 1 3
3 1 3 1 3
C B C B
E1,E2 << E3 E2 ,E3 << E1 E1,E2 >> E3 E1 ,E3 >> E2
c) d)
0.7
oxidizing, reducing,
basic basic
0.6 1x x
ad
FAB
ad
0.5 FAB
oxidizing, reducing,
0.4 acidic acidic
x
PBeq 1
0.3 0.4 0.5 0.6
Peq
Figure 2
42 Stochastic Pumping
a) I1 I2 b) ∼
ω
=
0.1 Pss(ε(t))
EL
∼
ω
=
1
u F(u(t))
∼
μ2 ω
=
10
EA Fast
μ1
ε ψ(t)
PELss(ε) Slow
reservoir 1 reservoir 2 Slow
k1 k2
EB
Ll + E Fast
EL E + L2
k-1 k-2 F(u)
c) d)
normalized current
F area = constant
1 area ~ ω
0.55
0.6 S
0.4
0.2 ∆I*¦ dΡ
à
*,ss
I¦* = I¦ + âV
ÙS ∆I¦ â S V dt
-3 -2 -1 0 1 2 3 4 dΡ
∼)
log( ω = I¦
*,ss
+ F* à âV
V dt
Figure 3
ADP
L1 L2 L1 L1 L2
L2 L1 L2
ATP
EA ADP
EA EB EB
Figure 4
Astumian 43
U(x,ψ)
Ψ
A
1 2 1 2
When on 1
A B 1 x 2
U(x,ψ)
Ψ Ψ A
1 2 1 2 1 x 2
A B B
Maxwell’s Demon- Demons at rest - Smoluchowski’s Demon-
Introduces correlations detailed balance Destroys correlations
βψ
ke βaψ
k eβψ
1A βψ
2A 1A k e βbψ 2A 1A 2A
ke ke k
γ γ γ e2β(b-a)ψ γ γ γ γ γ γ γ
−βψ
1B ke
2B k e−βaψ k e−βψ
−βψ
1B −βbψ
2B 1B 2B
ke ke k
Clockwise cycling No cycling Clockwise cycling
[2] > [1] [2] = [1] [2] > [1]
Figure 5
d) Reciprocal “cycle”
a) Thermally activated binding +O2
+O2 Fe O2 Fe O2 Fe O2 Fe
Fe O2 Fe O2 Fe O2 Fe
-O2
O2 Fe Fe O2 Fe O2 Fe O2 Fe
Fe Fe
hν
c) Thermally activated dissociation
+O2
-O2 -O2
O2 Fe O2 Fe O2 Fe Fe
Fe Fe Fe
DeoxyMb
f)
Fe
Fe
hν, -O2
OxyMb
O2 Fe +O2
Fe
-O2
O2 Fe
O2 Fe
Figure 6
44 Stochastic Pumping
a) aW
Bound
EAL EBL
Free
EA EB
(1-a) W
b) exp{β[(UEAL-UEBL)-(1-a)W]} c) exp{β[(UEAL-UEBL)-(1-a)W]}
EAL EB L EAL EBL
L L L1 L2 L1 L2
EA EB EA EB
exp{β[(UEB-UEA)-a W]} exp{β[(UEB-UEA)-a W]}
d) Pumping e) Catalysis
[L] > Kd [L] < Kd [L1] > Kd > [L2]
EB
EB EAL EA
EA EBL +L1
+L -L
EAL EAL
EB
EB L
EBL EA -L2
EB
EA
Figure 7