LETTERS

PUBLISHED ONLINE: 15 NOVEMBER 2009 | DOI: 10.1038/NMAT2566

Gel-expanded to gel-condensed transition in

neurofilament networks revealed by direct
force measurements
Roy Beck1 *, Joanna Deek2 , Jayna B. Jones1 and Cyrus R. Safinya1 *

Neurofilaments (NF)—the principal cytoskeletal constituent of there were no direct measurements of forces and interactions in
myelinated axons in vertebrates—consist of three molecular- neurofilament network gels.
weight subunit proteins NF-L (low), NF-M (medium) and Here, we address the underlying role of distinct side arms,
NF-H (high), assembled to form mature filaments with and the effects of altered subunit composition and salinity in
protruding unstructured C-terminus side arms1–5 . Liquid- modulating the forces between neurofilaments in their isotropic
crystal gel networks of side-arm-mediated neurofilament and liquid-crystal gel phases. Neurofilament subunit proteins
assemblies have a key role in the mechanical stability of NF-L (62 kDa), NF-M (103 kDa) and NF-H (117 kDa) have a
neuronal processes. Disruptions of the neurofilament network, common structure comprising a short unstructured N-terminus
owing to neurofilament over-accumulation or incorrect side- head domain followed by a hydrophobic α-helical body region
arm interactions, are a hallmark of motor-neuron diseases and an unstructured C-terminus side-arm tail domain (Fig. 1a).
including amyotrophic lateral sclerosis3–9 . Using synchrotron The subunit proteins share a conserved core (residues 93–412)
X-ray scattering, we report on a direct measurement of forces in with more than 65% sequence homology. The largest source of
reconstituted neurofilament gels under osmotic pressure (P). sequence variation occurs in the side arms. In this region, they
With increasing pressure near physiological salt and average differ in total protein length, net charge and charge distribution
phosphorylation conditions, NF-LMH, comprising the three (Supplementary Table S1 and Fig. S1). The side arms are overall
subunits near in vivo composition, or NF-LH gels, undergo negatively charged polyampholytes. In addition, NF-M and NF-H
for P > Pc ≈ 10 kPa, an abrupt non-reversible gel-expanded to have 26 and 49 residues respectively, which on phosphorylation
gel-condensed transition. The transition indicates side-arm- add −1.8 e charge to the neutral residues near physiological pH
mediated attractions between neurofilaments consistent with (Supplementary Table S1).
an electrostatic model of interpenetrating chains. In contrast, The assembly into mature neurofilaments begins with the
NF-LM gels remain in a collapsed state for P < Pc and transition formation of coiled-coil dimers of the hydrophobic α-helical
to the gel-condensed state at P > Pc . These findings, which regions of the subunits, which pair, in a half-staggered manner,
delineate the distinct roles of NF-M and NF-H in regulating to form tetramers (Fig. 1b). In cross-section, the neurofilament
neurofilament interactions, shed light on possible mechanisms is composed on average of four paired tetramers. Whereas NF-L
for disruptions of optimal mechanical network properties. will dimerize with another NF-L, the other subunits NF-M
The cell mechanical stability and elastic properties are primarily and NF-H require NF-L to form dimers1,2,10,13,21 . The assembled
determined by hierarchical cytoskeletal structures resulting from neurofilament consists of a 10 nm semiflexible backbone with
interactions among the constituent proteins. In long myelinated the unstructured C-terminus side arms sprouting radially from
neuronal axons, the highly abundant neurofilaments form an the core fibre1,4,17,18,21–23 . Representative atomic force microscopy
aligned nematic liquid-crystal gel with an open-network struc- (AFM) images of neurofilaments are shown in Fig. 1c–e, where
ture, which acts as a structural scaffold. In addition to rigidity the typical flexible necklace-like structure is clearly shown1,10,13,16,24 .
under external pressures, it is clear that transport of organelles Whereas the side arms are invisible by the AFM technique, each
along microtubules immersed within the neurofilament gel will bead is associated with the α-helical-rich region spaced by the
require a flexible and well-dispersed network9–11 . Indeed, aber- linker heads1,10,13,24 .
rations in neurofilament network structure, including incorrect To explore the nature of the interaction between neurofilaments
phosphorylation and deviations in subunit composition leading to and determine the role of the different neurofilament subunits in
abnormal accumulation of neurofilament aggregates, are hallmarks stabilizing the neurofilament network, we carried out a synchrotron
of motor-neuron diseases such as amyotrophic lateral sclerosis, small-angle X-ray scattering (SAXS) study of neurofilament gels
Lewy-body-based dementia and Parkinson’s3–9,12 . subjected to osmotic pressure (P) at different subunit compositions
Experiments involving electron microscopy imaging9,10,13,14 and and salt concentrations (see the Methods section and refs 25–27).
rheology15–19 have focused on the nature of crossbridging be- This technique allowed us to directly measure the interactions
tween side arms and its effect on the network properties. between neurofilaments and their subunit proteins and to provide
In addition, a recent phase-behaviour study is consistent with direct experimental evidence for pressure-induced attractions (that
interactions that are dependent on salinity and side-arm com- is, crowding) in neurofilament gels. The proteins were purified from
position in neurofilament gels20 . Nevertheless, before this study bovine spinal cord and reassembled in the desired composition.

1 Materials,
Physics, and Molecular, Cellular, and Developmental Biology Departments, University of California Santa Barbara, California 93106, USA,
2 Chemistry
and Biochemistry Department, University of California Santa Barbara, California 93106, USA. *e-mail: roy@mrl.ucsb.edu;
safinya@mrl.ucsb.edu.

40 NATURE MATERIALS | VOL 9 | JANUARY 2010 | www.nature.com/naturematerials

© 2010 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT2566 LETTERS
a b

NF-L N
C

NF-M N
C

NF-H N
C

Residue number 1 90 400 555 926 1,090

c d h

250 nm 500 nm

1.5 mm
1 µm

Figure 1 | Hierarchical neurofilament self-assembly. a, Neurofilament subunit protein schematics. The three subunits share a similar N-terminus globular
head domain and an α-helical body. The main difference between the proteins is found in the intrinsically unstructured C-terminus side-arm domain.
b, Single-filament self-assembly hierarchy. Composed of 32 subunit proteins in cross-section, neurofilaments self-assemble and form a flexible backbone,
10 nm in diameter (D), with radiating unstructured side arms. The side arms are half staggered, with 16 side arms every 22.5 nm (h) along the
neurofilament backbone. For NF-LM filaments (NF-L assembled with increasing amounts of NF-M), the maximum side-arm density is achieved at
(NF-L/NF-M 52:48 wt%) and corresponds to 11 NF-M and 21 NF-L subunits in cross-section. This results in protruding NF-M side arms separated by
≈5.7 nm around the circumference20 . Likewise, for NF-LH filaments, the maximum side-arm density (NF-L/NF-H 60:40 wt%) corresponds to eight NF-H
and 24 NF-L subunits in cross-section, with an NF-H side-arm separation of ≈7.8 nm around the circumference20 . c–e, AFM images show a necklace
structure and a preferred parallel configuration even at a low protein concentration. f, Cross-polarized microscopy image of a neurofilament network in a
quartz capillary. The nematic domain boundaries are a clear indication of the extended length scale of the ordered phase. The non-spherical bubble (yellow
arrow) reveals the yield stress resisting surface tension, as expected for a gel. The inset illustrates the possible configurations and microscopic interactions,
such as physical entanglement or intrafilament and interfilament interactions, leading to the nematic order.

Neurofilament gels consisted of homopolymers of NF-L, or NF-L The alignment of the neurofilament hydrogel is well
plus NF-M (NF-LM), NF-L plus NF-H (NF-LH) or all three demonstrated in cross-polarized microscopy, where millimetre-
NF-LMH mixtures near physiological composition. In all cases, sized domains are observed (Fig. 1f). Gelation of the neurofilament
neurofilament gel pellets phase-separate from the excess buffer network is confirmed by the appearance of rough surfaces at
after an initial swelling, demonstrating an inherent attractive the gel–buffer interface and by the occasional appearance of
interaction at low pressure20 . non-spherical bubbles inside the pellet (see the arrow in Fig. 1f).

NATURE MATERIALS | VOL 9 | JANUARY 2010 | www.nature.com/naturematerials 41

© 2010 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE MATERIALS DOI: 10.1038/NMAT2566

a b
103

102

(PEO wt%, P (kPa))

101
102
(20, 536) (PEO wt%, P (kPa))

(10, 107)
Intensity (a.u.)

Intensity (a.u.)
100 (15, 267)
(5, 26) (12.5, 175)
(2.5, 8) 101
(10, 107)
10¬1 (1.25, 3)
(2.5, 8)
(0.75, 1.6)
(1.25, 3)
10¬2 (0, 0.1)
100

(0, 0.1)
10¬3
0.01 0.10 0.01 0.10
q (Ŭ1) q (Ŭ1)

c d
104 102

(PEO wt%, P (kPa))

103 (PEO wt%, P (kPa))
(20, 536)

(17.5, 385)
Intensity (a.u.)

Intensity (a.u.)

101
102
(15, 267)
(12.5, 175)
(5, 26)
(10, 107)
101 (7.5, 59) (2.5, 8)

(5, 26) (1.25, 3)

(0.25, 0.6) 100 (0.5, 1)

100
(0, 0.1) (0, 0.1)

0.01 0.10 0.01 0.10

q (Ŭ1) q (Ŭ1)

Figure 2 | SAXS for different osmotic pressures and subunit compositions. Profiles of azimuthally averaged SAXS data of neurofilament networks under
different osmotic pressures and subunit protein compositions in 150 mM monovalent-salt buffer. a, SAXS profiles of neurofilaments reconstituted at an
NF-LMH 7:3:2 mol ratio, the physiological subunit ratio. b, SAXS profiles of NF-L homopolymer, representing the simplest system with identical
158-amino-acid-long side arms extending from the filament backbone. The contribution of each of the longer side arms has been explored through the
NF-LM and NF-LH binary systems, NF-L being a ubiquitous component as both NF-M and NF-H are NF-L-obligated heterodimers. c,d, SAXS profiles of
NF-LM (65:35 wt% ≈ 4:1 mol:mol) (c) and NF-LH (70:30 wt% ≈ 6:1 mol:mol) (d) copolymers. The weight percentage of PEO and its corresponding
osmotic pressure is noted. Lorenzian line-shape fits to the SAXS profiles (dotted line) were used to determine the peak positions. A power-law background
(dashed line) was added to the Lorenzian fit.

SAXS–osmotic pressure data were taken for a range of salt Figure 3 summarizes the pressure versus interfilament distance,
concentrations both near (90–240 mM) and far (40, 70 mM) d, obtained from the SAXS profiles for the neurofilament gels
from physiological ionic strengths. Figure 2 shows SAXS data studied for different subunit compositions and salt concentrations.
of NF-L, NF-LM, NF-LH and NF-LMH gels with increasing Filled (open) symbols represent nematic (isotropic) gels. Neurofil-
pressure in 150 mM salt buffer. The broad correlation peaks of aments comprising the NF-L homopolymer represent the minimal
the nematic structure factor (Fig. 2) are centred about wavevector system of a flexible backbone with 158 residue side arms decorating
q1 inversely proportional to the average interfilament distance it in a bottlebrush fashion. Interfilament spacing between NF-L
between neighbouring filaments d(= 2π/q1 ). As expected, the filaments is seen to vary slightly over four decades of pressure with
correlation peak positions, q1 , shift to higher values as the pressure d ranging between 36 and 46 nm at low pressure for monovalent-
is elevated and the gel network is compressed. salt concentrations between 40 and 240 mM (Fig. 3a). The weak

42 NATURE MATERIALS | VOL 9 | JANUARY 2010 | www.nature.com/naturematerials

© 2010 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT2566 LETTERS
a b
107 107
NG salt (mM)

40
106 106
70
90
105 105
150
P (Pa)

P (Pa)
240
104 IG, salt (mM) 104

40
103 90 103

240
102 102
200 400 600 800 200 400 600 600
d (Å) d (Å)

c d
107 107

106 106

105 105
P (Pa)

P (Pa)

104 104

103 103

102 102
200 400 600 800 200 400 600 800
d (Å) d (Å)

e P < Pc f P > Pc g

Figure 3 | Osmotic pressure–distance curves for neurofilaments of different subunit protein compositions and buffer salinities. a, NF-L homopolymer
is a basal system having the shortest negatively charged side arms. The pressure profiles of these homopolymers are weakly salt dependent and exemplify
a strong repulsive interaction between neighbouring side-arm coronas. b, NF-LMH (7:3:2 molar ratio) polymer represents a recombined neurofilament of
physiological subunit ratio. The network is expanded at low osmotic pressures and undergoes a transition at high pressure to a condensed state.
c, Filaments recombined with NF-L and NF-H (NF-LH, 70:30 wt% ≈ 6:1 mol:mol) show similar trends to those of NF-LMH (7:3:2 molar ratio), including the
high-osmotic-pressure transition. d, NF-LM (65:35 wt% ≈ 4:1 mol/mol) copolymer is in an expanded state at low monovalent-salt concentrations (40,
70 mM), and a collapsed state (g) at physiological salt concentrations. The X symbols in b,c represent a second, salt-independent, correlation peak that
follows the NF-L curve trend (dashed line). The lines drawn to the physiological salt pressure profiles are simply a guide to the eye. The filled and open
symbols in a–d represent the nematic gel (NG ) and isotropic gel (IG ) conditions respectively (information deduced from crossed-polarized microscopy on
the hydrogel samples). e–g, Cross-section schematics for expected side-arm interaction at expanded (e) condensed (f) and collapsed (g) gel states.
Whereas the condensed state will have abundant ionic bonds of interpenetrating chains, the collapsed state should have some transient interpenetration.

salt dependence and large interfilament spacing suggests that the the pressure is increased, the weakly aligned expanded-gel state
charged NF-L side arms and counter-ions form a nearly neutral is found to be highly incompressible, with d decreasing only
corona with stretched side-arm chains owing to the pressure of slightly with increasing pressure over nearly two decades before
the counter ions. the onset of an abrupt transition for P > Pc ≈ 10 kPa from large
The P–d profile is qualitatively different for neurofilament gels to small interfilament spacing, indicative of a gel-expanded to
comprising NF-LMH filaments near physiological composition gel-condensed transition (Fig. 3b,c,e,f). The P–d profile is weakly
(7:3:2 mol ratio for NF-L/NF-M/NF-H) (Fig. 3b) or NF-LH salt dependent and the transition at high pressure can be noted even
gels (Fig. 3c). We find that the average interfilament spacings at a monovalent-salt concentration of 240 mM (Fig. 3b,c).
at P < 10 kPa are nearly twice (≈70–80 nm) that of NF-L This abrupt transition signals the onset of an attractive
(≈36–45 nm) and the gels are in an expanded state (Fig. 3e). As interaction overcoming a repulsive one. The repulsion, which

NATURE MATERIALS | VOL 9 | JANUARY 2010 | www.nature.com/naturematerials 43

© 2010 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE MATERIALS DOI: 10.1038/NMAT2566

a d
550
20
N2 (residue number)

10
500 2 1

ΔE(kBT)
10 0

450 0

ΔE(kBT)
-10

¬10
-20
400 ¬20
400 450 500 550 535 540 545 550 555
N1 (residue number) N1 (residue number)

b e
900
N2 (residue number)

2 1

700

10
ΔE(kBT)

0
500
-10

500 700 900 400 500 600 700 800 900

N1 (residue number) N1 (residue number)

c f

1,000

2 1
N2 (residue number)

900

800
10
700
ΔE(kBT)

600
0
500

-10
600 800 1,000 800 850 900 950 1,000 1,050
N1 (residue number) N1 (residue number)

Figure 4 | Handshake analysis of complementary side-arm interactions. Two side arms aligned in an antiparallel configuration, showing the side-arm
interaction between neighbouring filaments. The colours in the 1E(N1 ,N2 ,w = 10,m = 10) handshake matrices, given by equation (1), represent the
Coulomb interaction energetics surrounding residues N1 and N2 from opposing side arms. The black markers represent potential sites for hydrophobic
interactions between the pair of opposite side arms using the Kyte and Doolittle method30 (see Supplementary Methods). a, The NF-L–NF-L handshake
matrix shows only a weak possible attraction that arises in the tip–tip and tip–body configurations. b, The NF-M–NF-M handshake matrix reveals
complementary charged regions at tip–body and configurations beyond full overlap. In addition, the matrix shows hydrophobic regions capable of
interaction in the tip–tip and tip–body configurations. c, The NF-H–NF-H handshake matrix lacks any hydrophobic regions but contains many potential
attractive electrostatic regions along the body of the side arms. d–f, The dashed lines in a–c represent the corresponding configurations (shown in d–f),
where the magnitude of these interactions is found to be of the order of a few kB T.

dominates in the gel-expanded state, is most likely due to transitions26 . Moreover, we find that the attractive interaction
undulating neurofilaments comprising overall negatively charged leading to the transition at P > Pc is irreversible over the
side arms23,28 . This behaviour is qualitatively similar to other timescale of a month, as the interfilament spacing only mildly
biological systems undergoing condensing transitions, for example, increased after replacing the poly(ethylene oxide) (PEO)-rich
the liquid-expanded to liquid-condensed29 and DNA-bundling with PEO-free buffer. This irreversibility implies interpenetration

44 NATURE MATERIALS | VOL 9 | JANUARY 2010 | www.nature.com/naturematerials

© 2010 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT2566 LETTERS
of opposing side arms in the gel-condensed state (Fig. 3f and point out that the repulsive electrostatic interaction between most
Supplementary Fig. S2), in contrast to a model of non-penetrating of the body regions of opposing NF-L side arms (Supplementary
compressed chains. Fig. S1), together with the high NF-L side-arm density along
For NF-LH and NF-LMH gels, the SAXS data best fit two the filament core, presents a strong repulsive barrier against
correlation peaks. The extra peaks (X symbols in Fig. 3), at interpenetration of chains and is consistent with our observation
low interfilament spacing, follow the same trend as the NF-L that NF-L filaments remain repulsive even at high pressure.
homopolymer (dashed line). This suggests either coexistence of At similar levels of pH, ionic strengths and the average
NF-L gels and NF-LH gels (resulting from the self-assembly process phosphorylation state, the side-arm interactions for NF-M and
when the two protein subunits are mixed) or coexistence within a NF-H are qualitatively different. For NF-M side arms (Fig. 4b),
filament of regions of pure short NF-L side arms and regions of the although the tip–body regions show some favourable interactions,
extended NF-H side arm. The latter case would not be surprising as the strongest attractions are between the side-arm residues near
the density of NF-H side arms (1:6 NF-H/NF-L subunits) is low. the neurofilament core with the nearby residues of the opposing
For salt concentrations near physiological conditions side arm (diagram in Fig. 4e, with side arms extending beyond full
(90–240 mM), NF-LM gels show P–d force curves similar to NF-L overlap). These regions would be accessible (for example, when
gels with slightly larger average interfilament spacings (Fig. 3d). neurofilament cores cross each other, see Supplementary Fig. S2)
This is indicative of a collapsed transition at P < Pc , shown at high salt concentrations, which would effectively screen the large
schematically in Fig. 3g, where NF-M side arms may show few repulsive interactions between residues along the body regions
transient chain interpenetrations, and a transition into a condensed of opposing side arms (that is, the central yellow/red-coloured
state for P > Pc . Interestingly, NF-LM filaments do indeed show regions, Fig. 4b), or by means of dephosphorylation that neutralizes
the expanded-gel state behaviour at P < Pc and the transition to negatively charged residues . This is consistent with Fig. 3d, where
the gel-condensed state at P > Pc , but only at low ionic strengths NF-LM gels for P < Pc are collapsed with allowable transient chain
(40–70 mM) below typical physiological ionic strengths. This can be interpenetration (Fig. 3g) at high salt concentrations and expanded
seen for a high composition of NF-M in NF-LM filaments (Fig. 3d, (Fig. 3e) for low ionic strengths.
open squares and circles). At high salt concentrations, P–d curves For NF-H side arms, one finds that although the tip–body
for all NF-LM gels collapse to the NF-L homopolymer curve. interactions found for NF-L side arms are also found here, more
Although the net charge of the side arm is negative, the significantly, NF-H side arms show many favourable interactions
polyampholyte nature of the side arms allows matching of (each of the order of a few kB T) scattered all along the body regions
positive and negative charge residues in opposite side arms. To of opposing side arms for interpenetrating configurations (that is,
systematically evaluate the possibility of attractive interactions the green/blue-coloured regions, Fig. 4c,f). The main favourable
between the side arms, we considered a possible ‘handshaking’ interaction sites for NF-H side arms are for partially overlapped
between opposite residues owing to short-range electrostatic side-arm configurations (that is, a more extended state of the
interactions. In this simple model, two side arms from neighbouring chains), whereas those for NF-M side arms are for side arms
filaments are aligned in an antiparallel configuration separated extending beyond full overlap, which leads to a more collapsed
by approximately 0.28 nm (Fig. 4, Supplementary Text and Fig. state. This is consistent with our observation that for a given ionic
S3). The energy landscape between residues from opposing side strength NF-LH gels are in a more extended state than NF-LM gels.
arms represented as a two-dimensional matrix can be estimated Our experimental results show that for physiological ionic
from Coulomb’s law: strengths and P < Pc the neurofilament network interfilament
spacing, with average side-arm phosphorylation, may be tuned
w/2 X
X m
eZ1 (N1 + i)eZ2 N2 − i − j

over a large range by the relative amounts of NF-H and NF-
1E(N1 ,N2 ) = ke (1) M side arms within the mature neurofilament, with NF-H
i=−w/2 j=−m |r1 (N1 + i) − r2 N2 − i − j |


(NF-M) side arms favouring expanded (collapsed) states with
high (low) mechanical flexibility. Future studies should clarify
Here, ke is the electrostatic constant. For a given antiparallel whether enhanced phosphorylation of NF-M side arms may
configuration, N1 and N2 refer to the residue numbers facing each mediate the gel-expanded state in NF-LM gels at physiological salt
other on the two opposing side arms (where the maximum residue concentrations. Finally, these findings suggest that a sudden change
number is the C-terminus amino acid) and eZ1 (N),eZ2 (N) and in the pressure, externally or owing to crowding of other proteins
r1 (N), r2 (N) are the charge and location of residues within each in the cell, can induce a transition to the less flexible neurofilament
chain, respectively. For each residue, the local electrostatic energy gel-condensed state, irreversibly altering its elastic properties,
is calculated between the residue in one side arm and its near and which require flexibility for its many functions including, for
several next-nearest neighbours (up to a distance of the side-arm’s example, transport of organelles along microtubules immersed
persistence length ≈3.5 nm) in the opposite side arm. The number within the neurofilament gel.
of nearest neighbours included is (2m + 1) and the effective coarse-
grained interaction is summed over a window containing w amino Methods
Neurofilament sample preparation. The neurofilament subunit proteins were
acids. We calculated all possible opposite side-arms configuration isolated through a two-part procedure. The first part carried out according to refs 15,
matrices and find that indeed there are numerous configurations 20, begins with bovine spinal cord and produces a solution enriched in neurofila-
that allow partial handshaking between neurofilament side arms ments and depleted of cellular debris. As the crude purification does not involve the
(Fig. 4 and Supplementary Information). chemical disruption and/or denaturation of the single filamentous structure, the re-
The origin for the inherent attractive interaction between the sultant crude neurofilament solution will not only contain neurofilaments, but also
other associated and co-isolating proteins. The second part involves the separation
subunit proteins may now be better understood. Our electrostatic under denaturation conditions (8 M urea) of the three subunits from all associated
analysis shows the energetically favourable gain for handshaking and co-isolating proteins, as well as from one another20 . Binary and ternary neuro-
owing to the polyampholyte nature of NF-L, NF-M and NF-H side filament subunit mixtures are exchanged into and assembled to filaments in buffer A
arms. For NF-L side-arm–side-arm interactions, we find favourable (100 mM MES, 1 mM MgCl2 , 1 mM EGTA, 3 mM NaN3 , at pH = 6.8). The filaments
interactions primarily between the C-terminus tip region of a side are sedimented to form neurofilament hydrogels by ultracentrifugation at 100,000 g
for 1 h at room temperature (Beckman Airfuge, Rotor A-10/30). The hydrogels are
arm with regions along the length of the opposing side arm as two then transferred to 1.5 mm quartz capillaries (Hilgenberg), and then capped with
opposing side arms explore overlapping configurations (Fig. 4a and 100 µl of buffer A. The salinity of buffer A was modified by adding KCl. All chemicals
schematic in Fig. 4d showing tip–tip and tip–body interactions). We were obtained from Sigma Aldrich and used without further purification.

NATURE MATERIALS | VOL 9 | JANUARY 2010 | www.nature.com/naturematerials 45

© 2010 Macmillan Publishers Limited. All rights reserved.
LETTERS NATURE MATERIALS DOI: 10.1038/NMAT2566

Osmotic pressure. We change the osmotic pressure by varying the concentration 16. Wagner, O. I. et al. Softness, strength and self-repair in intermediate filament
of 20,000-molecular-weight PEO. The conversion between the weight percentage of networks. Exp. Cell Res. 313, 2228–2235 (2007).
PEO in solution and its effective osmotic pressure on the hydrogel was done using 17. Gou, J. P., Gotow, T., Janmey, P. A. & Leterrier, J. F. Regulation of
the equation from Parsegian et al.25 : log(P) = 1.61 + [PEO wt%]0.21 , where P is the neurofilament interactions in vitro by natural and synthetic polypeptides
effective osmotic pressure in pascals. Furthermore, to account for the barometric sharing Lys–Ser–Pro sequences with the heavy neurofilament subunit NF-H:
pressure on the neurofilament hydrogel owing to the buffer column volume, 138 Pa Neurofilament crossbridging by antiparallel sidearm overlapping. Med. Biol.
is added to the total pressure calculated from the formula listed above. Eng. Comput. 36, 371–387 (1998).
18. Storm, C., Pastore, J. J., MacKintosh, F. C., Lubensky, T. C. & Janmey, P. A.
AFM. Diluted neurofilament (1–100 µg ml−1 ) solutions were deposited onto a Nonlinear elasticity in biological gels. Nature 435, 191–194 (2005).
clean poly(ethyleneimine)-coated Si wafer. After drying, the sample was measured 19. Rammensee, S., Janmey, P. & Bausch, A. Mechanical and structural properties
under ambient pressure in tapping mode and at 1 Hz/line speed using a Digital of in vitro neurofilament hydrogels. Eur. Biophys. J. 36, 661–668 (2007).
Instruments Nanoscope IIIa Dimension 3100 and an Asylum MFP-3D. 20. Jones, J. B. & Safinya, C. R. Interplay between liquid crystalline and isotropic
gels in self-assembled neurofilament networks. Biophys. J. 95, 823–835 (2008).
SAXS. Samples were measured in sealed quartz capillaries at the Stanford 21. Geisler, N. & Weber, K. Self-assembly in vitro of the 68,000 molecular
Synchrotron Radiation Lightsource and the Advanced Light Source (Berkeley). To weight component of the mammalian neurofilament triplet proteins into
avoid radiation damage, a test sample was measured before each run to determine intermediate-sized filaments. J. Mol. Biol. 151, 565–571 (1981).
the maximal exposure time. Subsequently, samples were irradiated for no longer 22. Heins, S. & Aebi, U. Making heads and tails of intermediate filament assembly,
than half that exposure time. The samples were stable for over six months at room dynamics and networks. Curr. Opin. Cell Biol. 6, 25–33 (1994).
temperature and were measured in triplicate over that time period. The correlation 23. Kumar, S., Yin, X. H., Trapp, B. D., Hoh, J. H. & Paulaitis, M. E. Relating
peaks from the azimuthally averaged SAXS data were fitted, after background interactions between neurofilaments to the structure of axonal neurofilament
subtraction with the form of A · q−B + C (B ∼ 2–3), to the Lorenzian function distributions through polymer brush models. Biophys. J. 82, 2360–2372 (2002).
using Matlab routines. 24. Hesse, H. C. et al. Direct imaging of aligned neurofilament networks assembled
using in situ dialysis in microchannels. Langmuir 24, 8397–8401 (2008).
Received 22 June 2009; accepted 5 October 2009; 25. Parsegian, V. A., Rand, R. P., Fuller, N. L. & Rau, D. C. Osmotic-stress for
published online 15 November 2009 the direct measurement of intermolecular forces. Methods Enzymol. 127,
400–416 (1986).
References 26. Rau, D. C. & Parsegian, V. A. Direct measurement of the intermolecular forces
1. Janmey, P. A., Leterrier, J. F. & Herrmann, H. Assembly and structure of between counterion-condensed DNA double helices—evidence for long-range
neurofilaments. Curr. Opin. Colloid Interface Sci. 8, 40–47 (2003). attractive hydration forces. Biophys. J. 61, 246–259 (1992).
2. Xu, Z. S. et al. Subunit composition of neurofilaments specifies axonal 27. Needleman, D. J. et al. Synchrotron X-ray diffraction study of microtubules
diameter. J. Cell Biol. 133, 1061–1069 (1996). buckling and bundling under osmotic stress: A probe of interprotofilament
3. Xu, Z., Cork, L. C., Griffin, J. W. & Cleveland, D. W. Increased expression of interactions. Phys. Rev. Lett. 93, 198101–198104 (2004).
neurofilament subunit NF-L produces morphological alterations that resemble 28. Zhulina, E. B. & Leermakers, F. A. M. A self-consistent field analysis
the pathology of human motor neuron disease. Cell 73, 23–33 (1993). of the neurofilament brush with amino-acid resolution. Biophys. J. 93,
4. Fuchs, E. & Cleveland, D. W. A structural scaffolding of intermediate filaments 1421–1430 (2007).
in health and disease. Science 279, 514–519 (1998). 29. Pallas, N. R. & Pethica, B. A. Liquid-expanded to liquid-condensed transitions
5. Miller, C. C. J. et al. Axonal transport of neurofilaments in normal and disease in lipid monolayers at the air water interface. Langmuir 1, 509–513 (1985).
states. Cell. Mol. Life Sci. 59, 323–330 (2002). 30. Kyte, J. & Doolittle, R. F. A simple method for displaying the hydropathic
6. Liu, Q. et al. Neurofilament proteins in neurodegenerative diseases. Cell. Mol. character of a protein. J. Mol. Biol. 157, 105–132 (1982).
Life Sci. 61, 3057–3075 (2004).
7. Brady, S. Mice overexpressing the human neurofilament heavy gene as a model Acknowledgements
of ALS. Neurobiol. Aging 16, 490–492 (1995). We thank R. Bruinsma, M. Kardar, Y. Li, A. Parsegian, P. Pincus, R. Podgornik, M.
8. Hirano, A., Donnenfeld, H., Sasaki, S. & Nakano, I. Fine-structural observations Wyrsta and E. Zhulina for discussion and correspondence and M. C. Choi for help
of neurofilamentous changes in amyotrophic lateral sclerosis. J. Neuropathol. with the AFM studies. The X-ray research was carried out at the Stanford Synchrotron
Exp. Neurol. 43, 461–470 (1984). Radiation Lightsource and at the Lawrence Berkeley National Laboratory Light Source.
9. Nguyen, M. D. et al. A NUDEL-dependent mechanism of neurofilament AFM experiments were conducted in the Microscopy & Microanalysis Facility of the
assembly regulates the integrity of CNS neurons. Nature Cell Biol. 6, UCSB Materials Research Laboratory (supported by NSF-MRSEC). C.R.S., J.D. and J.B.J.
595–608 (2004). were supported by DOE BES DE-FG-02-06ER46314 (protein purification and assembly,
10. Hirokawa, N. in The Neuronal Cytoskeleton (ed. Burgoyne, R. D.) 5–74 characterization) and NSF DMR-0803103 (phase behaviour). R.B. was supported by the
(Wiley–Liss, 1991). Human Frontier Science Program organization and in part by DOE BES.
11. Hirokawa, N. Organelle transport along microtubules—the role of KIFs.
Trends Cell Biol. 6, 135–141 (1996). Author contributions
12. Brownlees, J. et al. Charcot–Marie–Tooth disease neurofilament mutations This letter was written by R.B., C.R.S. and J.D.; R.B. collected and analysed the X-ray,
disrupt neurofilament assembly and axonal transport. Human Mol. Genet. 11, AFM and microscopy data and assisted in purifying the protein; J.D. purified the
2837–2844 (2002). proteins and collected the X-ray and microscopy data. J.B.J. carried out preliminary
13. Hirokawa, N., Glicksman, M. & Willard, M. Organization of mammalian measurements and refined the protein purification protocol. C.R.S. designed the original
neurofilament polypeptides within the neuronal cytoskeleton. J. Cell Biol. 98, study; R.B., J.D. and J.B.J. were involved in the study design. The data were interpreted
1523–1536 (1984). by R.B., J.D. and C.R.S.; J.B.J. commented on the final manuscript.
14. Brown, H. G. & Hoh, J. H. Entropic exclusion by neurofilament sidearms:
A mechanism for maintaining interfilament spacing. Biochemistry 36, Additional information
15035–15040 (1997). The authors declare no competing financial interests. Supplementary information
15. Leterrier, J. F., Kas, J., Hartwig, J., Vegners, R. & Janmey, P. A. Mechanical accompanies this paper on www.nature.com/naturematerials. Reprints and permissions
effects of neurofilament cross-bridges. Modulation by phosphorylation, lipids, information is available online at http://npg.nature.com/reprintsandpermissions.
and interactions with F-actin. J. Biol. Chem. 271, 15687–15694 (1996). Correspondence and requests for materials should be addressed to R.B. or C.R.S.

46 NATURE MATERIALS | VOL 9 | JANUARY 2010 | www.nature.com/naturematerials

© 2010 Macmillan Publishers Limited. All rights reserved.