Copyright © 2006 American Scientific Publishers Journal of

All rights reserved Computational and Theoretical Nanoscience

Printed in the United States of America Vol. 3, 1–14, 2006

Buckling, Bundling, and Pattern Formation:

From Semi-Flexible Polymers to Assemblies of
Interacting Filaments
Jan Kierfeld∗ , Petra Gutjahr, Torsten Kühne, Pavel Kraikivski, and Reinhard Lipowsky
Max Planck Institute of Colloids and Interfaces, Science Park Golm, 14424 Potsdam, Germany

Semiflexible polymers and filaments play an important role in biological and chemical physics. Single
filaments are characterized by a certain bending rigidity which governs their persistence length and
buckling instabilities. Attractive mutual interactions of filaments in bundles or the attractive interaction
with an adhesive substrate lead to equilibrium phase transitions, such as bundling and adsorption.
Finally, on the level of active systems consisting of many interacting filaments, we discuss coopera-
tive ordering effects in non-equilibrium systems such as motility assays. In motility assays filaments
are adsorbed and driven by motor proteins, which are anchored to a planar two-dimensional sub-
strate. The interaction with motor proteins leads to enhanced ordering of filaments. Motility assays
containing patterns of adsorbed motors, i.e., stripes of low motor density with increasing widths can
be used to sort filaments according to their lengths.
Keywords: Filaments, Semiflexible Polymers, Molecular Motors, Buckling, Bundling, Pattern
Formation.

RESEARCH ARTICLE
1. INTRODUCTION the renormalization of bending rigidity and the buckling
instability. The bending rigidity renormalization leads to
Stiff, filamentous polymers play an important role in bio- an effective softening of large-scale fluctuations of a fila-
logical and chemical physics. Both DNA or cytoskele- ment, which can be quantified using renormalization group
tal filaments such as F-actin and microtubules1
2 and methods.4 The buckling instability is an important effect
chemically synthesized stiff polymers such as dendronized that restricts the force generation by growing filaments.
polymers3 are “nanorods.” These polymers have diameters We discuss a cusp-like singularity in the projected length
in the range from 2 to 25 nanometers which leads to a as a function of the filament length, which can serve as
considerable bending rigidity and gives rise to persistence an experimental signature for the buckling of growing fil-
lengths comparable to or larger than the polymer’s contour aments or rods.
lengths. Such semiflexible polymers exhibit a variety of Then we will discuss the equilibrium phase transition
cooperative phenomena, which we will briefly review in that leads to the formation of filament bundles in the
this article. The cooperative behavior arises from the com- presence of attractive interactions, which can arise from
petition of several energies in the system, i.e., the bending crosslinking proteins or unspecific interactions.5 In eukary-
energy, the thermal energy, interaction energies, and exter- otic cells, the most important building blocks of the cyto-
nal driving forces. In biological systems, driving forces can skeleton are microtubules and filamentous actin (F-actin).
arise from the activity of molecular motors which perform Actin filaments have a persistence length Lp  30 m,6
directed motion on cytoskeletal filaments or the polymer- microtubules are much stiffer with a persistence length
ization dynamics of cytoskeletal filaments. Both the activ- Lp  10 mm.7 In the cortex of the cell, actin filaments
ity of molecular motors and the polymerization dynamics form a dense meshwork which is responsible for many
are processes far from equilibrium and are typically cou- of the viscoelastic properties of the cell. Another impor-
pled to the hydrolysis of adenine triphosphate (ATP). tant morphology that is found in the cell are filament
First, we will consider two aspects of single filaments bundles,8 which, e.g., support cell protrusions and serve as
related to the cooperative behavior of their bending modes, stress fibres. Both meshworks and bundles are hold together
by different actin-binding crosslinking proteins.8
9 Actin
∗
Author to whom correspondence should be addressed. bundling crosslinkers possess two adhesive end domains

J. Comput. Theor. Nanosci. 2006, Vol. 3, No. 6 1546-198X/2006/3/001/014 doi:10.1166/jctn.2006.006 1

 Buckling, Bundling, and Pattern Formation Kierfeld et al.

which bind to filaments by weak bonds; crosslinker medi- heads, and polymeric motor tails as separate degrees of
ated interactions therefore allow a reversible formation of freedom.
actin bundles, which can be regulated by the concentra- In such motility assays, several interesting ordering
tion of crosslinkers in solution. Solution of actin filaments phenomena can occur on the nanoscale if we consider
and crosslinking proteins have been studied in vitro in a many interacting filaments. The ordering is caused by the
number of recent experiments.10–12 In these studies it has interplay of the active dynamics of filaments and their
been observed that bundle formation in F-actin solutions mutual repulsive interactions. On the one hand, we find an
containing crosslinking molecules requires a threshold increased tendency for nematic ordering in motility assays
crosslinker concentration above which (i) F-actin bundles with randomly adsorbed motor proteins.25 On the other
become stable against the thermal fluctuations of filaments hand, stripe-shaped patterns of adsorbed motors can be
and (ii) a phase containing networks of bundles separates. used for the length sorting or fractionation of filaments.
Polymerizing bundles are used within the cell in order to This article is organized as follows. In Section 2, the
generate forces.13
14 We will shortly discuss one possible persistence length of a single filament is discussed in the
“zipping”-mechanism for force generation which is based framework of a systematic renormalization group approach.
on the conversion of adhesive energy into polymerizing In Section 3, we discuss a particular feature of the buck-
forces. ling instability of a single filament, which is relevant to
Another important equilibrium phase transition of poly- experiments on growing filaments. In Section 4, the forma-
mers is their adsorption to an attractive planar surface. tion of filament bundles via crosslinker-mediated attractive
For semiflexible polymers or filaments, the adsorption interactions and the ability of “zipping” bundles to gen-
transition is similar to the binding of two filaments but erate forces is discussed. The adsorption of a filament on
represents a distinct universality class.15 Various single an adhesive surface, which is provided by a planar two-
molecule methods have been applied to adsorbed semiflex- dimensional substrate covered with anchored motor pro-
ible polymers because both visualization and manipulation teins and which represents the geometry used in motility
are easier for adsorbed polymers with a large diameter, assays, is considered in Section 5. In Section 6, we intro-
such as DNA.16
17 These polymers are generically semi- duce a model for the active filament dynamics in motility
flexible because strong entropic or enthalpic interactions assays and present recent simulation results. These results
along their backbone increase the bending rigidity. The demonstrate an enhanced nematic ordering of interacting
manipulation kinetics of strongly adsorbed single filaments filaments in motility assay. They also demonstrate that fil-
or semiflexible polymers on structured substrates requires aments can be sorted according to their length if striped
RESEARCH ARTICLE

thermal activation.18
19 Here, we will discuss the adsorp- patterns of different motor density are realized in a motil-
tion behavior of filaments in motility assays. In such an ity assay.
assay, cytoskeletal filaments are adsorbed and driven over a
two-dimensional, planar substrate by motor proteins whose
tails are anchored to the substrate.20 In order to obtain 2. PERSISTENCE LENGTH
adsorption, a critical density of motor proteins is needed The coupling between the different bending modes of sin-
in close analogy to the critical crosslinker concentration gle filaments can lead to a considerable softening of the
for the formation of a filament bundle. filaments on large length scales. On length scales larger
Motility or gliding assays are a standard biochem- than the persistence length Lp , the polymer appears to be
istry assay to characterize motor proteins, which is based flexible. We have quantified this effect using renormaliza-
on measuring the lateral displacement of adsorbed fil- tion group (RG) methods; here we briefly summarize some
aments. In biological cells, small forces generated by of the results, details are presented elsewhere.4 Starting
motor proteins organize and rearrange cytoskeletal fila- from a discretized model for an inextensible semiflexible
ments and give rise to active, non-equilibrium filament chain consisting of M bonds or segments of length b and
dynamics, which plays an important role for cell divi- with unit tangent vectors ti (i = 1
  
M) in d spatial
sion, motility, and force generation.1
2 Whereas conven- dimensions. The bending energy of this semiflexible chain
tional “passive” polymer dynamics is governed by thermal is given by26
fluctuations,21 active filament dynamics is characterized by
a constant supply of mechanical energy by motor pro- 
M

teins, which hydrolyze ATP. Motility assays are model

ti  = 1 − ti · ti−1 
with ti2 = 1 (1)
b i=1
systems, which allow to study active filament dynamics
in a controlled manner. By analyzing the transport veloci- where  is the bending rigidity. Physically relevant cases
ties of single filaments gliding over the substrate, informa- are chains in three spatial dimensions (d = 3) and chains,
tion can be obtained about basic properties of molecular which are adsorbed on a planar substrate and, thus, con-
motors such as their maximal velocity. We introduce a fined to two dimensions (d = 2). In the continuum limit of
simulation model for motility assays, which refines previ- small b the model (1) reduces to the continuous worm-
ous models22–24 and contains semiflexible filaments, motor like chain model for an inextensible filament.27

2 J. Comput. Theor. Nanosci. 3, 1–14, 2006

Kierfeld et al. Buckling, Bundling, and Pattern Formation

The partition sum is obtained by integrating over all 3. BUCKLING INSTABILITY

possible bond orientations with the appropriate Boltzmann
weight exp −

ti /T , where T is the temperature in Another hallmark of filaments or nanorods with bending
energy units. Within the model (1), modes of different elasticity is the presence of buckling instabilities if the fil-
wavelengths are coupled. Upon integrating out short-scale ament is subject to a large compressional force of size F
fluctuations, this leads to a softening of the effective as follows from classical elasticity theory.30 In the absence
renormalized   governing the fluctuations with wave of thermal fluctuations, the filament undergoes such an
length  such that   < . In order to calculate  , instability if the compressional force F exceeds a certain
we use a real-space functional renormalization group threshold value, the critical force Fc , for constant filament
(RG) approach, which has been developed originally in length or if the filament length L exceeds a certain critical
the context of one-dimensional classical Heisenberg spin length Lc for constant force. We will limit our discussion
models.28 The RG transformation eliminates fluctuations to the situation of a longitudinal force F that acts parallel
up to the scale  = 2R b . After the Rth RG step, we can to the straight, unbuckled state of the filament. The thresh-
find an exact recursion relation for M =  2R b . For old values for force and filament length are then given by
d = 2, the resulting recursion relations are given by4 2
Fc = cbc  2 /L2 and Lc = cbc  /F 1/2 (5)
    
R 2R
0 2R 2

0 2R
=  K0  m m K0  where cbc is a dimensionless coefficient that depends on
 K0 m=− m m=− the boundary conditions at the two filament ends. For
with K0 ≡ /b T and  0
m K0  ≡ e
−K0
Im K0  the three boundary conditions (bc0), (bc1), and (bc2)
(2) described in the next paragraph and illustrated in Figure 2,
one has cbc = 1/2
1, and 2, respectively.
where Im x denotes the modified Bessel function of the Let us denote the two ends of the filament by e1 and e2 .
first kind.29 A similar recursion relation can be found for Filament end e1 is taken to be immobilized and to have a
d = 3. The renormalized   for d = 2 and d = 3 is plot- fixed spatial position. The orientation of the filament seg-
ted in Figure 1 as a function of the length scale . For ment adjacent to e1 , which is described by the tangent
large , the expression (2) has the asymptotic behavior vector of this segment, may be free or clamped: ‘free’
  T means that it can freely adapt to the compressional force,
≈  2e−T /2 − 4e−T / + 8e−3T /2 − · · ·  (3) ‘clamped’ that it is constrained to be parallel to this force.
 
The three boundary conditions are now defined as follows

RESEARCH ARTICLE
which is governed, to leading order, by the decay length
(see also Fig. 2): (bc0) Filament end e1 is clamped and
Lp = 2/T which we identify with the persistence length
filament end e2 is free. In this case, a longitudinal force
in d = 2. Similarly, we define Lp = /T in d = 3. These
that acts parallel to the straight, unbuckled filament leads
results generalize the conventional definitions based on the
to a displacement of e2 that has both a longitudinal and a
exponential decay of the particular two-point tangent cor-
transverse component; (bc1) Both filament ends are free.
relation function and gives identical results for Lp . In the
As the filament buckles, the filament end e2 is only dis-
remainder of this article we will define the persistence
placed in the longitudinal direction parallel to the force;
length by the d = 2 result, i.e.,
and (bc2) Both filament ends are clamped and we only
Lp ≡ 2/T (4)
(bc0) (bc1) (bc2)
F F F F

1 e2 e2 e2

L|| L||

e1 e1 e1
κ( )/κ

0.5
Fig. 2. Straight unbuckled filament and three boundary conditions
(bc0), (bc1), and (bc2) for buckling. (Left) Straight, unbuckled fila-
ment growing against the upper (grey) wall which exerts the longitudinal
force F . The projected length L
of the filament is equal to its contour
length L. The straight filament is stable as long as L do not exceed a criti-
cal value Lc , which differs for the three boundary conditions (bc0), (bc1),
and (bc2), see Eq. (5). For L > Lc the filament buckles and L
< L. The
5000 10000
lower filament end e1 has a fixed spatial position. For (bc0), filament end
/ b e1 is clamped to an orientation parallel to the force and filament end e2
is free; for (bc1), both filaments ends can freely adapt their orientations;
Fig. 1.  / as a function of /b = 2R for K0 = 1000 for d = 2 for (bc2), both filament ends have clamped orientation and, furthermore,
(dashed line) according to the recursion relation (2 and d = 3 (solid line)). we only allow for displacements of e2 parallel to the force.

J. Comput. Theor. Nanosci. 3, 1–14, 2006 3

 Buckling, Bundling, and Pattern Formation Kierfeld et al.

allow for a longitudinal displacement of e2 parallel to the 1

force.
In the living cell, the ATP-driven polymerization of fil- 0.8
aments is an important mechanism for force generation.
Likewise, synthetic semiflexible polymers, which grow via 0.6

1– L || /Lc
polymerization, can be envisaged as suitable systems for
force generation on the nanoscale. The ability of these sys- 0.4
tems to perform work is, however, limited by the buckling
instability. Here we want to discuss a simple model sys- 0.2

tem where a single filament is growing by polymerization

against a planar wall. As before, one filament end, e1 , is –0.75 –0.5 –0.25 0 0.25 0.5 0.75 1
immobilized and has a fixed spatial position. The other
L/Lc – 1
end, e2 , can now grow by attachment of monomers which
leads to a time-dependent filament length L = L t. Since Fig. 3. Plot of the reduced projected length L
/Lc as a function of
this growth is rather slow, we can assume that the fila- the reduced contour length L/Lc for the two boundary conditions (bc0)
ment attains a state of mechanical equilibrium with length and (bc1), see text. For L < Lc , the filament is straight with L
= L
which corresponds to the left part of the diagram with L/Lc − 1 < 0.
L = L t. The planar wall is movable and is loaded by a
The buckled solution appears for L > Lc as shown in the right part of
constant force F , such that it always touches and pushes the diagram with L/Lc − 1 > 0. The red curve is obtained numerically
the growing end of the filament. Now we investigate how by a parametric plot using ∗ = L as the curve parameter. The black
the buckling instability of the filament is reflected in the line is the linear approximation (11). For L/Lc − 1 > 0478, L
becomes
behavior of the projected length L
, which is the longitu- negative.
dinal extension of the filament in the direction parallel to
the force and is directly accessible in experiments. and  y 1 − cos x
If the filament bends in a plane, its configuration can 2 y ≡ dx  (10)
2 cos x − cos y
0
be described by the planar angle s between its tangent
vector and the force direction which is parametrized by As y goes to zero, the first integral has the finite limit
the arc length s with 0 < s < L. The total energy E of the 1 0 = /2 whereas 2 0 = 0. Using these expressions,
filament then has the form one can obtain a parametric representation of the reduced
 L   projected length L
/Lc as a function of the reduced con-
RESEARCH ARTICLE

E= ds #s 2 + F cos s (6) tour length L/Lc in the buckled state with L > Lc . For the
0 2 two boundary conditions (bc0) and (bc1), one obtains
where the first and the second term represent the bending the same relation which is plotted in Figure 3. Close to
and compression energy, respectively. The force F is pos- the buckling instability, all three boundary conditions lead
itive if it acts against the growing filament. The projected to the asymptotic behavior
length L
is given by 1−L
/Lc ≈ 3 L/Lc −1 for small L/Lc −1 > 0 (11)
L
L
= ds cos s (7) For L < Lc , on the other hand, the filament is unbuckled
0 which implies that the projected length L
is identical with
Minimizing the total energy E with respect to the angle the contour length L and
configuration s, leads to the beam equation which has 1 − L
/Lc = 1 − L/Lc for L/Lc − 1 < 0 (12)
to be solved for appropriate boundary conditions. This
solution can be expressed in terms of the planar angle Combining the two results for L > Lc and L < Lc , we
see that the relation between projected and contour length
∗ ≡ s∗ . For the boundary conditions (bc0) and (bc1),
one simply has s∗ = L. For the boundary condition (bc2), exhibits a cusp at the buckling point with L = Lc as shown
the arc length s∗ corresponds to the smallest s-value for in Figure 3.
which the filament’s curvature vanishes which implies For a growing filament this cusp-like singularity directly
d s/ds = 0 for s = s∗ . One then finds that the ratios of translates into a singularity in time. The filament growth
the contour length L and of the projected length L
to the can be characterized by the rate %eff F  of subunit inser-
critical length Lc are given by tion or attachment which depends on the external force F .
The time-dependent length of the filament is then given by
L   L
1 ∗  − 2 ∗
= 1 ∗ and = (8) L t = Lc + b %eff F  t − tc  (13)
Lc 1 0 Lc 1 0
with the two integrals where b is the size of the inserted subunits and tc is the
buckling time. As an example, let us consider an actin
y 1
1 y ≡ dx  (9) filament which is extended by a formin molecule with
0 2 cos x − cos y insertion rate %eff 0  25 subunits/s and subunit size

4 J. Comput. Theor. Nanosci. 3, 1–14, 2006

Kierfeld et al. Buckling, Bundling, and Pattern Formation

b  275 nm.31 Using the growth law (13) for L t in (11)

for t > tc and in (12) for t < tc , one obtains the jump

dL
dL

− = −4 b %eff F  (14)
dt t=tc +0 dt t=tc −0 z2 (x) z1 (x)
x

in the growth rate dL

/dt of the projected length at the Fig. 5. Crosslinkers connecting two filaments. If these crosslinker
buckling instability. This singularity should be accessible molecules are rather rigid, they mediate effective filament–filament inter-
to experiments and can then be used to determine the actions that depend on the filament orientations as discussed in Ref. [15].
growth rate %eff F  of the filament or nanorod. The singu- In this review, we will focus on the case of short, flexible crosslinkers
larity as given by (14) also applies to N parallel filaments with two adhesive endgroups as indicated in this figure. In the latter case,
the effective filament–filament interactions depend only on the distance
provided one replaces F by F /N .
between the filaments.

4. FILAMENT BUNDLES with 0 < x < L

, where L
is the projected length of
the polymer. This parametrization is appropriate provided
Within this section we consider an equilibrium phase tran- the longitudinal correlation length of the displacements is
sition in a system of many filaments, which have a short- small compared to Lp . The filament interaction is mediated
range attractive interaction mediated by crosslinkers. We by crosslinking sticker molecules that adsorb from the sur-
consider N filaments with bending rigidity  in a solu- rounding solution. Each linker consists of a short polymer
tion containing crosslinking molecules with two adhesive with adhesive endgroups, see Figure 5. In the following,
end groups. This system exhibits a critical crosslinker we assume flexible linkers for simplicity which can link
concentration, X1 = X1
c , which separates two different two filaments irrespective of their orientation; orientation-
concentration regimes. For X1 < X1
c , the filaments are dependent interactions which arise from rigid linkers can
unbound and uniformly distributed within the compart- be studied in a similar way, see Ref. [15].
ment. For X1 > X1
c , the filaments form either a single We discretize the filament into segments of length a
,
bundle, which represents the true ground state of the sys- i.e., xk = ka
and zi
k = zi xk . The presence or absence
tem as in Figures 4(a) and (c), or several sub-bundles, of a crosslinker molecule at segment k of filament i
which represent metastable, kinetically trapped states as in is described by the occupation number ni
k = ni xk  =
Figure 4(b). Furthermore, as we decrease the crosslinker 0
1. The filament-crosslinker system is governed by the
concentration from a value above X1
c towards a value Hamiltonian

RESEARCH ARTICLE
below X1
c , the bundles undergo a discontinuous unbind-


ing transition at X1 = X1
c . The existence of a single,
= 
b
i
zi  +
1
ni  +
2
zi − zj
ni
nj  (15)
i i
j
discontinuous unbundling transition can be established by
analytic methods for N = 2 filaments15 and by Monte where the first contribution
L 1
Carlo (MC) simulations for larger bundles containing up

b
i = dx  #x2 zi 2
to N = 20 filaments.5 0 2
contains the bending energies of the filaments. The term
4.1. Model
1 describes the intrafilament interactions of linkers. We
consider a lattice gas of linkers with hard-core repulsion
The filaments are oriented along one axis, say the x-axis, adsorbing on a filament with
and can be parametrized by two-dimensional displace-

ments zi x (i = 1
  
N ) perpendicular to the x-axis,
1 = a
Wni
k
k

(a) (b) (c) where W < 0 is the adhesive energy (per length) of one
linker end group. The third contribution
2 describes the
pairwise interactions between filaments i and j and is
given by
 

1

2 = a
Vr /zij
k + ni
k +nj
k −2ni
k nj
k Va /zij
k 
k 2
(16)
where /zij
k ≡ zi
k − zj
k . The first term represents the
hard-core repulsion between two filaments that is indepen-
dent of the linker occupation and leads to a potential
Fig. 4. Monte Carlo snapshots of bundles with N = 20 filaments.
(a) Close to the unbinding transition in the bundled phase. (b) Deep in Vr z =  for z < r and Vr z = 0 otherwise
the bound phase, the bundle tends to segregate due to slow kinetics and
filament entanglement. (c) The equilibrium shape of the bundle is roughly The length scale r is of the order of the filament diameter.
cylindrical. The second term is the linker-mediated attraction and is
J. Comput. Theor. Nanosci. 3, 1–14, 2006 5
 Buckling, Bundling, and Pattern Formation Kierfeld et al.

non-zero if either segment k of filament i or segment k (a) 2

of filament j carries a linker. Then the other filament is 0
attracted by a linker-mediated potential Va z. The latter
filament gains the additional energy W  if /zij
k  ≤ a , –2
where the potential range a is of the order of the linker –4
N=2

size. This attraction is modeled by the potential well

<H>/N L||
a

–6
Va z = W for 0 < z−r < a and Va z = 0 otherwise N=3
–8
(17) b
N = 20
We can perform the partial trace over the crosslinker –10 N = 10
N=5
degrees of freedom ni
k in the grand-canonical ensem- –12
ble to obtain an effective interaction between filaments. N = 10
N = 20
c
Each crosslinker has two adhesive ends. The first adhe- –14
4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9
sive end adsorbs on a filament and establishes the standard |W|
Langmuir-type adsorption equilibrium with a linker con- (b) 1
centration per site X1 ≡ ni
k 1 = cx / Kd + cx  where the N=2 N = 20
average is taken with the Hamiltonian
1 . X1 is thus deter-
mined by the concentration cx of linkers in solution, where N = 10

Kd is the equilibrium constant of the dissociation reaction

of the crosslinker with the filament. Tracing over weakly N = 20
<∆z> N = 10
bound linkers with W   T /a
, we end up with effective 0.1
pairwise linker-mediated filament interactions, i.e.,
N=3
N=5
1


2 ≈ a
Vr /zij
k  + Va /zij
k 
2 k N=3

which have the same functional form as the bare interac- N=2

tions; the short-range attractive part Va is of the form (17)

0.01
0.01 0.1 1
with a strength† (|W|–|Wc|)/|W|
 ≈ 2X1 W
RESEARCH ARTICLE

W (18)
Fig. 6. MC data for N = 2
3
5
10
20 identical filaments (with per-
proportional to the linker concentration on the filament. sistence length Lp = 200, contour length L
= 100, potential range a =
Pairwise filament interactions with potentials of the form 0001, and hard core radius r = 01; all lengths are in units of /x,
(17) are generic and do not only arise from crosslinkers but lines are guides to the eye). For N = 10
20 two branches of data are
shown corresponding to two different initial conditions; in the lower
also from van-der-Waals, electrostatic, or depletion forces. branch we prepared a compact cylindrical configuration, in the upper
branch (thick lines) we arranged filaments initially in a plane. (a) Mean
4.2. Discontinuous Unbundling Transition energy
/NL
per filament (in units of T ) as a function of the effec-
tive potential strength W  (in units of T //x). Arrows correspond to the
We have studied bundle formation by MC simulations for snapshots in Figure 4. (b) Logarithmic plot of the mean filament sepa-
ration /z ≡ /zij  − r  (in units of /x) as a function of the reduced
up to N = 20 identical filaments (i = ) using the effec-   − W  .
c /W
potential strength W
tive Hamiltonian


directly proportional to the mean bundle thickness that can

=
b
i +
2 be determined by optical microscopy in experiments.
i i
j
Our MC simulations confirm that, for bundles con-
Filaments are discretized into L
//x points along the taining up to N = 20 filaments, there is a single, dis-
x-direction, in which we apply periodic boundary condi- continuous unbinding transition, see Figure 6(a). In the
tions. In each MC step we attempt a random perpendicular presence of a hard-core repulsion, the critical potential
strength W c saturates to a N -independent limiting value
displacement in the z-direction. The MC simulations can
Wc
 for large N . The numerical simulations support a
be used to determine the locus and order of the unbind-
scaling behavior W c ∼ N − with an exponent
c
 − W
ing transitions, at which the mean energy
 exhibits
a discontinuity, see Figure 6(a). To gain further insight   10 ± 01 as shown in Figure 7. As can be seen in
into bundle morphologies, we also measure the mean seg- Figure 4(a) typical bundle morphologies close to the tran-
ment separation /zij  − r , see Figure 6(b), which is sition are governed by pair contacts of filaments. The bun-
dle thickness, as given by the mean segment separation
†
This corrects the corresponding Eqs. in Ref. [5]; The effective inter- /zij  − r , stays finite up to the transition, see MC data
action
2 contains an additional factor 1/2, and the effective potential in Figure 6(b). For increasing N , an increasing number
strength W  an additional factor 2. of higher moments /zij  − r m  remains finite at the

6 J. Comput. Theor. Nanosci. 3, 1–14, 2006

Kierfeld et al. Buckling, Bundling, and Pattern Formation

F
5.6

5.4
∆x
5.2
Wc

4.8
∆x
4.6

4.48
0 1/10 1/5 1/3 1/2
x
1/N

Fig. 7. MC data points for the critical potential strength Wc  as a func-

tion of the inverse filament number 1/N . The data points are fitted well
c
  448.
by a straight line with a limiting value W

transition (all moments m < 2 N − 1 3N − 4/3 remain

finite) showing that the critical thickness fluctuations of
large bundles become small.
 , our MC Fig. 8. Schematic illustration of the zipping mechanism. Upon zipping
Deep in the bundled phase, i.e., for large W
or bundling of filaments over an additional length /x they gain an adhe-
simulations show that bundles do not always reach their sion energy J/x. Motion of the wall over the same distance /x against
equilibrium shape. Small sub-bundles containing typically an external force f requires an energy f/x.
N ∼ 5 filaments form easily, start to entangle, and fur-
ther equilibration is kinetically trapped suggesting that the rod system separates into a high-density nematic phase
bundle is in a “glass” phase. Figure 4(b) shows the seg- and a low-density nematic or isotropic phase above a crit-
regation into sub-bundles in a typical configuration and ical attraction, which is in qualitative agreement with the
Figure 6(a) shows the corresponding rise in the mean bun- experimental results in Refs. [10–12].
dle energy per filament which approaches the N = 5 result.

RESEARCH ARTICLE
In Figure 6(b) the pronounced rise of the mean separa- 4.3. Force-Generation by Bundling or Zipping
tion for N > 5 with increasing potential strength and with
increasing N is due to the segregation. This behavior is The adhesive energy which is gained during bundle forma-
reminiscent of the experimentally observed F-actin struc- tion can be used to generate forces. The basic mechanism
tures consisting of networks of small bundles.11 Only when can be explained for two filaments which are bundling
starting from a sufficiently compact initial state, bundles or zipping due to a short-range attraction as discussed in
relax towards the equilibrium form in the MC simulation, the previous section. As sketched in Figure 8, we start
which is a roughly cylindrical bundle with a hexagonal with a bundle of two filaments which is oriented in the
filament arrangement as shown in Figure 4(c). In contrast x-direction. One end of the bundle—the lower end in
to the segregated form, the bundle thickness and the mean Figure 8—is spatially fixed and has a clamped orienta-
energy per filament of the equilibrium form decrease with tion. At the other end of the bundle—the upper end in
increasing N , as can be seen in Figure 6. Figure 8—the two filaments unbind and bend against a
The critical potential strength Wc corresponds to a criti- rigid wall. The wall exerts a total force F in the negative
cal crosslinker concentration X1
c . For weakly bound link- x-direction, such that each filament is loaded by a force
ers W   T /a
, we have a simple linear relation W  ≈ F /2. As the filaments bind or zip together over an addi-
2X1 W such that X1
c ≈ W c /2W . The corresponding rela- tional length /x, they gain a free energy J/x, where J > 0
tion for strongly bound linkers is more complicated but is the free energy of adhesion (per length) as introduced
X1
c will increase monotonically with increasing W c . in the previous section. In the absence of thermal shape
Our simulations use periodic boundary conditions and fluctuations of the filaments, the free energy is equal to the
treat very long and essentially parallel filaments. In order energy of adhesion J ≈ W  . In the presence of thermal
to include translational and rotational entropy we can map fluctuations, this ‘bare’ adhesion energy is considerably
the ensemble of semiflexible filaments considered here reduced by entropic contributions until it finally vanishes
onto an ensemble of rigid rods of finite length L and diam- as J ∼ W c − W close to the unbundling transition as dis-
eter a⊥ at a certain concentration c. The effective pair- cussed above. Zipping together an additional length /x
wise attraction (per length) J is given by the bundling free implies that the wall moves by the same distance /x which
energy of the filaments with J ∼ W  > 0 for W
c − W  > costs an energy F /x. The total zipping free energy gain is

Wc . Using the results of Ref. [32], we find that the hard Ezip = J − F /x and zipping happens spontaneously for

J. Comput. Theor. Nanosci. 3, 1–14, 2006 7

 Buckling, Bundling, and Pattern Formation Kierfeld et al.

forces f that satisfy (a)

F < Fzip = J (19)

where Fzip is the maximal force which can be generated

by this simple mechanism. Deep in the bundled phase,
we have
 

Fzip = J ≈ W    W
for W c  (20)

The zipping mechanism does not rely on polymerization

forces but requires a length reservoir for the filaments in
front of the wall. This length reservoir has to be con-
stantly renewed by polymerization at the free polymer ends (b) 0.6
as long as the zipping proceeds. Force generation by the
zipping mechanism is eventually limited by the buckling 0.4
instability of the bundled zipped stem. Similar zipping
mechanisms have been proposed to play an essential role 0.2 growth

in the motility of nematode sperm cells.33
34

We have performed MC simulations of the zipping
vw
0
mechanism for two identical filaments using an effective
Hamiltonian


0.2
shrinking

=
b
i +
2
0.4
i i
j

which we parameterize in terms of the contour length L 0.6

of the filaments. Each filament is discretized into L//s 0
|Wc|
20 40
|W|
60 80 100

equidistant points along its contour. Initially, we prepare a F

bundle oriented in the x-direction parallel to the direction
Fig. 9. (a) Snapshot from MC simulations of zipping for two polymers
of the force with one end fixed and clamped, and the other with Lp = 2000 (in units of /s), an adhesive potential of W   = 50, and
end in a splayed configuration as shown in the snapshot a force F = 50 (both in units of T //s), i.e., close to the critical force
RESEARCH ARTICLE

Figure 9 on the left. During the MC simulation additional Fzip . The wall is visible in dark grey, the immobilized clamped ends of
monomers can attach at the splayed ends (gaining a bind- the filaments are outside the range of the picture in the front. (b) Mean
ing energy Em < 0), which provides the necessary reservoir velocity vw  (in units of /s per MC time step) of the pushing wall for
an adhesive potential with W   = 50 as a function of the pushing force
in polymer length. The wall is loaded with a force F and F . The other parameters are as in Figure 6; the critical potential strength
moved instantaneously (or adiabatically) to the monomer for bundling is W c   32, i.e., the simulations are performed deep in the
position with the largest x-coordinate after each MC step. bundled phase. The arrow corresponds to the snapshot on the left. For
After a steady state has established for a given force F we F < Fzip we find vw  > 0, and the zipped part of the bundle is growing.
measure the mean velocity vw  of the pushing wall in the For F > Fzip , it shrinks and vw  < 0. The MC data is consistent with
Fzip ≈ W  = 50, see (20).
positive x-direction. In the MC simulations we find indeed
a threshold behavior with a critical force Fzip in agreement
with our argument above. At low pushing forces F < Fzip , 5. FILAMENT ADSORPTION
we find vw  > 0, and the zipped part of the bundle is
constantly growing, whereas at high forces F > Fzip , it con- The adsorption transition of a single filament onto a pla-
stantly shrinks since vw  < 0. Our MC data are consistent nar substrate is qualitatively similar to the bundle for-
with the above estimate (20) for W    Wc , i.e., deep in mation for N = 2 filaments in 1 + 1 dimensions, where
the bundled phase, see Figure 9 on the right. the one-dimensional perpendicular distance z x from
For bundles consisting of N filaments we expect the free the surface is analogous to a one-dimensional separation
energy gain from binding or zipping N filaments together between filaments. The adsorption transition can be solved
over an additional length /x to scale as NqJ/x/2 for large analytically,15 which reveals that unbinding and desorption
N , where q is the number of nearest neighbors: Because represent two distinct universality classes with different
of the repulsive part of their interaction, filaments in large critical exponents.
bundles only interact with a limited number q of nearest Here we want to consider the adsorption of a filament
neighbors for short-range attractions. Moving the wall by with persistence length Lp = 2/T on a planar two-dimen-
the same distance /x costs an energy F /x and, thus, we sional substrate where molecular motors are adsorbed
expect Fzip ∼ NqJ /2 for the maximal force that can be with an areal density 7. Each motor can bind to a fil-
generated by the zipping of a bundle consisting of a large ament within a capture radius m and a binding energy
number N of filaments. Wm < 0. Then each motor gives rise to an adsorption

8 J. Comput. Theor. Nanosci. 3, 1–14, 2006

Kierfeld et al. Buckling, Bundling, and Pattern Formation

potential Vad z of the same functional form as the The weak bending approximation is valid as long as gradi-
potential (17), ents are small, i.e., #x z2  ∼ 9
/Lp
1, which is fulfilled
ad − W
for W ad
c   T /Lp , which typically applies to stiff
Vad z = Wm for 0 < z < m and Vad z = 0 otherwise filaments such as microtubules adsorbed by kinesins.
(21)
In the following we assume that the motor tail is flex-
ible such that the filament can bind at any orientation. 6. MOTILITY ASSAYS FOR
A rigid motor tail eventually gives rise to an orientation- MOTOR PROTEINS
dependent adsorption potential because filaments could
Motility assays are model systems, which allow to study
bind only at a preferred angle. In Ref. [15] both cases have
active filament dynamics in a controlled manner. In motility
been discussed, and it has been found that the orientation-
assays, filaments are connected to the substrate by anchored
dependence can affect the order of the adsorption
motors of sufficient density 7 > 7c . In the presence of
transition. Whereas we find for orientation-independent
ATP, the motor heads start to perform a directed walk on
adsorption potentials a second order transition, the adsorp-
the filaments, which induces active dynamics of adsorbed
tion transition can become first order if the potential is
filaments. By analyzing the transport velocities of single
orientation-dependent.‡ In contrast to the case of the
filaments gliding over the substrate, information can be
annealed crosslinker ensemble considered previously, the
obtained about basic properties of molecular motors such
motors represent a partially quenched ensemble of adsorp-
as their maximal velocity. If many interacting filaments are
tion points. In the following, we consider the typical exper-
studied interesting active ordering phenomena can occur.
imental situation of a rather uniform coverage with motor
proteins and also neglect effects from filament fluctuations
parallel to the surface. Then the array of motors gives rise 6.1. Model
to an average adsorption potential Vad z of the same func- Our microscopic model for motility assays describes fil-
tional form as the potential (21) with a potential strength aments, motor heads, and polymeric motor tails as sepa-
Wad = Wm 7m . On length scales comparable or smaller
rate degrees of freedom.25 One end of the motor tail is
than Lp , the semiflexible polymer is only weakly bent by anchored to the substrate and the motor head on the other
thermal fluctuations and its configurations are governed by end can bind to a filament with the correct orientation
the effective Hamiltonian since the motor tail is rather flexible. Once bound the
 L
  
motor head moves along the filament thereby stretching

RESEARCH ARTICLE

ad = 2 2 
dx #x z + Vad z x (22)
0 2 the polymeric tail, which gives rise to a loading force act-
ing both on the motor head and the attached filament. This
We consider the limit of long filaments L
W ad  T , force feeds back onto the motion of the bound motor head,
which can exhibit a desorption transition. Using the model which moves with a load-dependent motor velocity.35
36
(22), this desorption transition has been studied by trans- Filaments follow an overdamped dynamics with external
fer matrix techniques in Ref. [15]. The critical potential forces arising from the stretched motor tails and the repul-
strength for desorption is W ad
c = −cT −2/3
m L−1/3
p corre- sive filament–filament interaction.
sponding to a critical motor density To proceed, let us consider N rigid filaments of length L
T on a planar two-dimensional substrate.¶ The configuration
7c = c (23) of filament i (i = 1
  
N ) is then specified by the posi-
Wm 5/3
m Lp
1/3
tion of its center of mass ri and its orientation angle ;i ,
√
where c ≈ 3/2 ≈ 15. For motor densities above the see Figure 10. The overdamped translational and rotational
critical density, filaments adsorb onto the substrate with dynamics of each filament i is described by the stochastic
anchored motors against the thermal fluctuations of fila- Langevin-type equations of motion25
ments. The critical motor density for adsorption is decreas- Ni



N
ing with increasing filament rigidity . Using estimates
· #t ri = Fi< Fij −  i (24)
Wm  15T and m  10−2 m for kinesin and Lp  10 mm <=1 j=1
for microtubules, we find a critical motor density 7c  10 and
Ni
m−2 for adsorption.


N

The transfer matrix treatment shows that the free energy =; #t ;i = Mi< + Mij + >;
i (25)
<=1 j=1
difference between adsorbed and unbound state vanishes
as /f  ≈ W ad
c w/ ln w−1 where w ≡ W ad − where Ni is the number of motor heads attached to fila-
 
Wad
c /Wad
c . Therefore, the correlation length 9
= ment i and indexed by <. The vector ui = cos ;i
sin ;i  is
T //f  ∝ w−: diverges with an exponent : = 1 + log.
¶
The model can be extended to deformable filaments by modeling each
‡
The case of an orientation-independent adsorption potential (21) cor- filament as a set of Ns segments connected by elastic springs and hinges,
responds to the case / = 0 in Ref. [15]. see P. Kraikivski, Ph.D. Thesis, Universität Potsdam (2005).

J. Comput. Theor. Nanosci. 3, 1–14, 2006 9

 Buckling, Bundling, and Pattern Formation Kierfeld et al.

ui motor tail at r0< and the filament is smaller than the cap-
α ture radius m . Apart from the stall force Fst the motor is
i riα also characterized by its detachment force Fd , above which
θi
ri
r0α the unbinding rate of the motor head becomes large. For
simplicity we assume in our model that the motor head
detaches whenever the force Fi< exceeds a threshold value
Fig. 10. Schematic view onto substrate surface: A filament i attached
Fd . We consider the case of processive motors with a high
by two motors attached. The vector ri points towards the filament’s center duty ratio close to unity, i.e., motors detach from a fila-
of mass, the parameters ;i and ui represent the orientational angle and ment only if they reach the filament end or if the force F
unit vector, respectively. The motor < is anchored at r0< , and its head is becomes larger than the detachment force Fd .
positioned at ri< .

6.2. Simulation
the orientational unit vector of filament i. Friction forces
are governed by the matrix of translational friction coef- Using the above model we can perform simulations of
ficients,
= =
ui ⊗ ui + =⊥ I − ui ⊗ ui ,21 where I is the gliding assays for different distributions of motors, i.e.,
unit matrix and =; is the rotational friction coefficient. random distributions or patterns of motors. We will first
All friction coefficients =
, =⊥ , and =; are known from focus on a random distribution of motors with a surface
passive filament dynamics. The translational and angular density 7 and a system with periodic boundary condi-
thermal random forces  i t and >;
i t, respectively, are tions. At each time step /t, we update the motor head
Gaussian distributed with correlations  i t ⊗  j t   = position xi< and filament position by using the discrete
2T
?ij ? t − t   and >;
i t>;
j t   = 2T =; ?ij ? t − t  . version of the equations of motion (25) and (26). The
The stretched tail motor < is exerting a force Fi< onto the parameter values that we choose for the simulations are
filament. The end-to-end vector of the polymeric tail is comparable with experimental data on assays for conven-
/r< ≡ ri< − r0< , where the motor tail is anchored at r0< and tional kinesin. The simulation results presented in this arti-
the head position is ri< . We model the polymeric tail as cle have been obtained for assays of quadratic geometry
freely jointed chain such that Fi< is pointing in the direc- and area 25 m2 with rigid filaments of length L = 1 m
tion −/r< and its absolute value is obtained by invert- and diameter D = L/40. We simulate at room tempera-
ing the force-extension relation of a freely jointed chain.26 ture T = 428 × 10−3 pN m. The friction coefficients are
There is also a corresponding torque due to the motor taken to be =⊥ = 2=
= 4AL/ ln L/D and =; = =
L2 /6,
RESEARCH ARTICLE

activity, Mi< =  ri< − ri  × Fi< . The interaction forces Fij where A is the viscosity of the surrounding liquid. We
and torques Mij are due to the purely repulsive interactions use a value A = 05 pN s/m2 much higher than the vis-
between filaments i and j corresponding to a hard-rod cosity of water, Awater ∼ 10−3 pN s/m2 , which allows to
interaction for filaments of diameter D. take larger time steps and decreases the simulation time.
The dynamics of motor heads is described by a deter- We checked that this does not affect results by performing
ministic equation of motion, which has the form selected simulation runs also at the viscosity of water. We
use a maximum motor speed of vmax = 1 m s−1 and a
#t xi< = v Fi<  (26) stall force of Fst = 5 pN. The capture radius for motor pro-
teins is m = 10−2 m and the length of the fully stretched
where xi<  ≤ L/2 defines the position of the motor < on motor tail Lm = 5 × 10−2 m.
the rod i, i.e., ri< = ri + xi< ui and the filament polarity is The motion of a single filament with contour length
such that the motor head moves in the direction ui . The L is characterized by stochastic switching between rota-
motor velocity v is a function of the loading force Fi< tional and translational diffusion if no motors are attached,
which builds up as the motor tail becomes more and more directed translation in rotationally diffusing directions if
stretched. We use a piecewise linear force-velocity relation one motor is attached, and directed translation in one
v Fi<  = vmax for Fi< ·ui ≥ 0 direction if two or more motors are attached. The relative

 frequency of these types of motion depends on the average
F<  number of motors attached to the filament or the average
= vmax 1− i for Fi< ·ui < 0 and Fi<  < Fst
Fst distance dm  between bound motors and, thus, on the sur-
=0 for Fi< ·ui < 0 and Fi<  > Fst (27) face motor concentration 7.22 In the limit of high motor
concentration a filament has two or more bound motors on
where vmax is the maximal motor velocity, which is average and dm  ∼ 1/7m . The single filament performs
attained if the motor is pulled forward. The motor speed a persistent walk with the effective persistence length22
decreases linearly if the motor is pulled backwards, until it

S;  L + 2 dm  L3 L
stalls if the backwards force exceeds the stall force Fst .35
36 9p = = e L/ dm 
− 1 −
We assume that the motor binds to the filament when /; 2 1/2 L + 3 dm  92m dm 
the distance between the position of the fixed end of the (28)

10 J. Comput. Theor. Nanosci. 3, 1–14, 2006

Kierfeld et al. Buckling, Bundling, and Pattern Formation

(a) 2500
7 Isotropic S < 0.2
2000 Nematic 0.2< S < 0.7
6 Nematic S >0.7
Sθ /
m

1500
5
1000 nematic

ρL2
500 4

3
100 200 300 400 500 (b) (c)
L/
m
2
(b)
1
isotropic
0
0.00 0.05 0.10 0.15
σ
mL

Fig. 12. The phase diagram of the gliding assay as a function of the
dimensionless filament density CL2 and dimensionless surface motor
density 7m L for a detachment force Fd = Fst and L/m = 100. Each
simulation run corresponds to one data point; arrows correspond to the
snapshots in Figure 11. If the average order parameter S < 02, the sys-
tem is in the isotropic phase (black squares, grey area), if S > 02 it is
(c) in the nematic phase (blue triangles, blue area). The solid line represents
the analytical result (30).

exhibits active nematic ordering if the motor density 7 is

increased as can be seen in the two simulation snapshots
Figures 11(b) and (c).25 Both the rod density C and the
motor density 7 are essential in order to determine the
phase behavior of the non-equilibrium motility assay. For

RESEARCH ARTICLE
given L/m , the corresponding phase diagram depends on
the dimensionless which can be described in the plane
of the two dimensionless parameters CL2 and 7m L as
shown in Figure 12. Nematic ordering in a system of N
Fig. 11. (a) Simulation results for average distance S;  traveled by a filaments can be characterized
 by the time averages of
filament between successive rotations as a function of the filament length the order parameter S ≡ i=j cos 2 ;i − ;j /N N − 1.
L for high motor concentration. The solid line is the analytical result (28) In equilibrium, i.e., in the absence of motors (7 = 0) we
as derived in Ref. [22]. (b) and (c) Snapshots of a gliding assay of rod- find a continuous isotropic-nematic transition at a critical
like filaments with filaments density C = 2/L2 on a motor coated substrate
with randomly distributed motors and periodic boundary conditions. For
density Cc
0  43/L2 in the simulation, which is in good
detachment forces Fd = Fst , we find (b) an isotropic phase at low motor agreement with the analytic result Cc = 3/2L2  47/L2
surface density 7m L = 003 and (c) active nematic ordering at high based on Onsager’s theory.37 The equilibrium transition is
motor surface density 7m L = 009. found numerically from the inversion point of the curve
where S;  is the average distance traveled by a filament S = S C for a value S  02, which we also use as
between successive rotations and /; 2 1/2 = 3/7L2 is the the threshold value for active nematic ordering if motors
average rotation angle at each rotation. The theoretical are present and 7 > 0, see Figure 12. Snapshots of the
result (28) is confirmed by our simulation as shown in actively driven system in the isotropic and nematic phase
Figure 11(a). The average filament velocity vF = ṙi  can are shown in Figures 11(b) and (c), respectively. In the
be obtained by simultaneously equating (i) the filament resulting phase diagram in Figure 12, the critical density
friction force with the total motor driving force and (ii) the Cc for active nematic ordering decreases with increasing
filament velocity with the motor velocity in the steady motor density, i.e., nematic ordering is favored if more
state, which leads to vF = vmax 1 + =
vmax dm /LFst −1 . mechanical energy is fed into the system. The transition
This relation is confirmed by our simulations. is continuous also for non-zero motor-density. In the pres-
ence of motor activity, there is a non-vanishing filament
6.3. Enhanced Ordering current in the nematic phase, which is characteristic for
a non-equilibrium phase. This filament current associated
Our results for the simulation of many filaments with with the nematic order breaks the rotational symmetry and
hard-core interactions indicate that the motility assay can be established only for periodic boundary conditions.

J. Comput. Theor. Nanosci. 3, 1–14, 2006 11

 Buckling, Bundling, and Pattern Formation Kierfeld et al.

For a closed system with hard wall boundary conditions, (a) Lα,1 Lα,2 Lα,3
on the other hand, we expect the formation of vortex-like
rotating filament patterns.
The simulation results can be explained using the con-
cept of an effectively increased filament length

Leff = L L + 9p  (29)

as compared to the equilibrium system, which explains

that motor activity actually favors nematic ordering. The
increased effective length is an effect of the persistent
motion of each filament, which effectively increases the
excluded volume area. This concept can be derived more
(b) Lα,1 Lα,2 Lα,3
rigorously by coarse-graining in time and averaging over
time intervals of one persistence time tp = 9p /vF in the
framework of a dynamical mean-field approach. We finally
obtain the critical filament density for the active isotropic-
nematic transition,
c
Cc = (30)
LL + 9p dm 
m
L
with c = 3/2 from the analytical mean-field calculation.
In the absence of motors we have 9p = 0 and the rela-
tion (30) reduces to the equilibrium result of Ref. [37].
Using the expression (28) for the persistence length 9p =
9p dm 
m
L, we obtain an explicit expression of the Fig. 13. Snapshots of gliding assays of a ternary mixture of rod-like
isotropic-nematic phase boundary in the active system in filaments of different lengths on a substrate coated with an alternating
pattern of stripes of low (71 = 0) and high (72 > 0) motor density. Each
terms of the microscopic model parameters, which is in
system of area 963 m2 contains a ternary mixture of N = 120 filaments
good agreement with the simulation data, see Figure 12. with 40 filaments of length L1 = 05 m, L2 = 1 m, and L3 = 14 m
RESEARCH ARTICLE

Beyond mean-field, we expect an increased numerical each, which corresponds to a total filament density of CL22 = 125. The
prefactor c in (30) but the same parameter dependence. stripes have increasing width from left to right with L7
1 /L1 = 058,
For thermally fluctuating filaments the phase diagram in L7
2 /L1 = 1165, and L7
3 /L1 = 173. The motor density 72 outside the
stripes is high with 72 m L2 = 024 for the assay (a) and 72 m L2 = 03
Figure 12 should be truncated for 7 < 7c , see Eq. (23),
for (b). Both snapshots clearly indicate the sorting of filaments according
where filaments can undergo thermal desorption from the to their lengths within the different stripes.
planar substrate.
their lengths by patterning the substrate with low motor
6.4. Filament Sorting density stripes which are arranged with increasing width
as shown in Figure 13 for three stripes. In general we can
An interesting application of motility assays is filament size sort filaments into N7 stripes with widths L7
1 < L7
2 <
sorting or fractionation on the nanoscale using substrates · · · < L7
N7 . A filaments of length L can bridge all narrow
coated with particularly designed patterns of molecular stripes until it encounters a stripe s with L7
s−1 < L < L7
s ,
motor density. Using simple gradient patterns this possibil- which is sufficiently broad to “trap” the filament. Figure 13
ity has already been explored experimentally in Ref. [38]. demonstrates that this sorting principle works for an assay
Here we propose a more efficient geometry and give a first consisting of a ternary mixture of filaments.
proof of principle using our simulation model.
If there are two spatial regions I and II on the substrate, 7. CONCLUSION AND OUTLOOK
which are characterized by different motor densities 7I
and 7II with 7I < 7II , filaments tend to move to region I Semiflexible polymers and filaments, which are ubiquitous
of low motor density 7I . If the substrate contains an alter- both in natural biological systems and synthetic biomimetic
nating pattern of stripes of low and high motor density, chemical systems, are nanorods which are subject to ther-
as shown in Figure 13, the filaments accumulate in the mal fluctuations and external forces. In this review we cov-
stripes with low motor density, unless the width L7 of ered some of the cooperative phenomena that can arise
the stripes of low motor density is larger than the fila- from the presence of a large number of internal bending
ment length L. Then, filaments can “bridge” the stripes modes or by interactions in filament assemblies.
of low motor density. This behavior can be used to con- We discussed the persistence length of a single fila-
struct an assay for the sorting of filaments according to ment using a renormalization group approach and some

12 J. Comput. Theor. Nanosci. 3, 1–14, 2006

Kierfeld et al. Buckling, Bundling, and Pattern Formation

characteristics of the buckling instability of a single fila- Lc critical length for buckling
ment. Then we moved on to systems containing ensem- Leff effective filament length (29)
bles of filaments and presented simulations and analytical Lp persistence length (4)
results for bundling transitions, which can also be used L7 width of stripe of low motor density
to generate forces on the nanoscale by means of zipping M bond number
mechanisms. Finally we presented a simulation model for Mi< torque onto filament i from motor head <
motility assays where filaments are subject to active fluc- Mij filament interaction torques
tuations from motor proteins that are immobilized on a ni
k crosslinker occupation on site k of filament i
substrate. This active driving leads to an enhanced ten- N number of filaments in a bundle or assay
dency for nematic ordering. N7 number of stripes
In this review, we focused on thermally fluctuating bio- ri center of mass of filament i
logical filaments. Principles which are realized in nature r0< anchor point of tail of motor <
can guide the design of artificial or synthetic devices, ri< position of head of motor < on filament i
which fulfill similar tasks as their biological counterparts. R RG step number
One particularly interesting example might be force gen- C filament density
eration on the nanoscale, where nature uses polymerizing Cc critical filament density for ordering
or zipping forces which could be imitated in synthetic or s arc length
biomimetic polymer systems in the future. Another exam- S nematic order parameter
ple where experimental work has already started are size S;  mean distance traveled by a filament
sorting assays for the fractionation of filaments.38 For syn- between rotations
thetic metal nanorods, also quantum fluctuations might 7 motor density
play a role in future experiments (see Ref. [39] for an t tangent vector
example), which have not been discussed in this review. t time
tp persistence time of filament motion
Acknowledgments: This work was supported by the T temperature (in energy units)
EC Sixth Framework Programme (as part of the STREP ;i angle of filament i
Active Biomics contract No. NMP4-CT-2004-516989). ui orientation of filament i
vF mean filament velocity
vmax maximal motor velocity

RESEARCH ARTICLE
LIST OF SYMBOLS vw  mean zipping velocity
Va z linker-mediated potential (17)
cx crosslinker conc. in solution Vad z adsorption potential (21)
D filament diameter Vr z hard-core repulsion
dm  mean distance between bound motors W adhesive energy (per length) of one
/r< end-to-end vector of motor tail linker end group
 bending rigidity 
W effective potential strength (18)
 RG length scale Wm motor binding energy
b bond length ad
W effective adsorption potential strength
a range of linker potential c
W critical potential strength for bundling
m motor capture radius x 1-dim. coordinate parallel to filament
r hard core size xi< position of motor < on filament i
A viscosity of the surrounding liquid X1 crosslinker conc. per filament site
F compressional force X1
c critical crosslinker conc. per site for bundling
Fc critical force for buckling 9p persistence length of filament motion (28)
Fd detachment force z d − 1-dim. coordinate perp. to filament
Fi< loading force onto filament i from motor  i t translational thermal random force
head < >;
i t rotational thermal random torque
Fij filament interaction forces
Fst stall force
Fzip maximal zipping force References
=
, =⊥ filament friction coefficients 1. D. Bray, Cell Movements: From Molecules to Motility, Garland Pub-
=; rotational friction coefficient lishers (2001).
s tangent angles 2. J. Howard, Mechanics of Motor Proteins and the Cytoskeleton, Sin-
auer Associates, Inc. (2001).
J free energy of adhesion 3. C. Ecker, N. Severin, L. Shu, A. D. Schlüter, and J. P. Rabe, Macro-
L filament contour length molecules 37, 2484 (2004).
L
filament projected length 4. P. Gutjahr, R. Lipowsky, and J. Kierfeld (unpublished).

J. Comput. Theor. Nanosci. 3, 1–14, 2006 13

 Buckling, Bundling, and Pattern Formation Kierfeld et al.

5. J. Kierfeld, T. Kühne, and R. Lipowsky, Phys. Rev. Lett. 95, 038102 21. M. Doi and S. F. Edwards, The Theory of Polymer Dynamics,
(2005). Clarendon (1986).
6. A. Ott, M. Magnasco, A. Simon, and A. Libchaber, Phys. Rev. E 48, 22. T. Duke, T. E. Holy, and S. Leibler, Phys. Rev. Lett. 74, 330 (1995).
R1642 (1993); J. Käs, H. Strey, and E. Sackmann, Nature 368, 226 23. M. R. Faretta and B. Basetti, Europhys. Lett. 41, 689 (1998).
(1994). 24. F. Gibbons, J.-F. Chauwin, M. Despósito, and J. V. Joseé, Biophys. J.
7. F. Gittes, B. Mickey, J. Nettleton, and J. Howard, J. Cell Biol. 120, 80, 2515 (2001).
923 (1993). 25. P. Kraikivski, R. Lipowsky, and J. Kierfeld, Phys. Rev. Lett. 96,
8. J. R. Bartles, Curr. Opin. Cell Biol. 12, 72 (2000). 258103 (2006).
9. K. R. Ayscough, Curr. Opin. Cell Biol. 10, 102 (1998); S. J. Winder, 26. J. Kierfeld, O. Niamploy, V. Sa-yakanit, and R. Lipowsky, Eur.
Curr. Opin. Cell Biol. 15, 14 (2003). Phys. J. E 14, 17 (2004).
10. M. Tempel, G. Isenberg, and E. Sackmann, Phys. Rev. E 54, 1802 27. O. Kratky and G. Porod, Rec. Trav. Chim. 68, 1106 (1949).
(1996). 28. Th. Niemeijer and Th. W. Ruijgrok, Physica 81A, 427 (1975).
11. O. Pelletier, E. Pokidyshevam, L. S. Hirst, N. Bouxsein, Y. Li, and 29. M. Abramowitz and A. I. Stegun, Handbook of Mathematical Func-
C. R. Safinya, Phys. Rev. Lett. 91, 148102 (2003). tions. Natl. Bur. Stand. (1965).
12. M. L. Gardel, J. H. Shin, F. C. MacKintosh, L. Mahadevan, 30. L. D. Landau and E. M. Lifshitz, Theory of Elasticity, Pergamon
P. Matsudaira, and D. A. Weitz, Science 304, 1301 (2004). (1986).
13. D. Pantaloni, C. Le Clainche, and M.-F. Carlier, Science 292, 1502 31. S. Romero, C. Le Clainche, D. Didry, C. Egile, D. Pantaloni, and
(2001). M.-F. Carlier, Cell 119, 419 (2004).
14. A. Mogilner and G. Oster, Curr. Biol. 13, R721 (2003). 32. M. Warner and P. J. Flory, J. Chem. Phys. 73, 6327 (1980);
15. J. Kierfeld and R. Lipowsky, Europhys. Lett. 62, 285 (2003); A. R. Khokhlov and A. N. Semenov, J. Stat. Phys. 38, 161
J. Kierfeld and R. Lipowsky, J. Phys. A: Math. Gen. 38, L155 (1985).
(2005). 33. A. Mogilner and G. Oster, Science 302, 1340 (2003).
16. S. S. Sheiko and M. Möller, Chem. Rev. 101, 4099 (2001). 34. L. Miao, O. Vanderlinde, M. Stewart, and T. M. Roberts, Science
17. N. Severin, J. Barner, A. A. Kalachev, and J. P. Rabe, Nano Lett. 4, 302, 1405 (2003).
577 (2004). 35. C. M. Coppin, D. W. Pierce, L. Hsu, and R. D. Vale, Proc. Natl.
18. P. Kraikivski, R. Lipowsky, and J. Kierfeld, Europhys. Lett. 66, 763 Acad. Sci. USA 94, 8539 (1997).
(2004); P. Kraikivski, R. Lipowsky, and J. Kierfeld, Eur. Phys. J. E 36. S. M. Block, C. L. Asbury, J. W. Shaevitz, and M. J. Lang, Proc.
16, 319 (2005). Natl. Acad. Sci. USA 100, 2351 (2003).
19. P. Kraikivski, R. Lipowsky, and J. Kierfeld, Europhys. Lett. 71, 138 37. R. F. Kayser and H. J. Raveché, Phys. Rev. A 17, 2067 (1978).
(2005). 38. L. Ionov, M. Stamm, and S. Diez, Nano Lett. 5, 1910 (2005).
20. J. Scholey, Motility assays for motor proteins. Methods in Cell Biol- 39. H. B. Chan, V. A. Aksyuk, R. N. Kleiman, D. J. Bishop, and
ogy, Academic Press (1993), p. 39. F. Capasso, Science 291, 1941 (2001).
RESEARCH ARTICLE

Received: 26 May 2006. Accepted: 18 July 2006.

14 J. Comput. Theor. Nanosci. 3, 1–14, 2006