1
Nature
Multiple liquid crystal phases
of DNA at high concentrations
Michael W. Davidson
Teresa E. Strzelecka & Randolph L.Rill
Department of Chemistry
and Institute of Molecular Biophysics
The Florida State University
Tallahassee, Florida 32306-3006, USA
Fig. 1 Dependence of solid state
31
P NMR linewidth on DNA concentration suggest-
ing multiple phase transitions. DNA fragments of average length 146 base pairs
(~500 Å) and with a narrow distribution about this length were prepared by digestion
with micrococcal nuclease of calf-thymus chromatin previously depleted of histone
H1, and by subsequent deproteinization
16
. Short DNA molecules are studied as a
preliminary step to understanding the natural behavior of DNA at in vivo concentra-
tions. Such defined samples are useful because effects of relative molecular mass
heterogeneity are minimized and phase transitions are sharp and kinetically rapid.
Using DNA fragments of this size is appropriate because the driving forces are the
same for ordering of high axial ratio, rod-like molecules and long semi-flexible
polymers of identical composition
5,6
. Open circles: data taken at 30ºC; filled circles:
data taken at 50ºC. Spectra were obtained on non-spinning samples at a phosphorus
frequency of 61.3 MHz on the ‘Seminole’, an in-house constructed multinuclear
Fourier transform (FT) spectrometer equipped with wide-bore superconducting
solenoid and quadrature detection, modified for solid-state applications. Sweep
width was ±25 kHz, pulse repetition times were 3 s, and 600 scans were added for
each spectrum. Gated proton decoupling was applied during data acquisition.
DNA packaging in vivo is very
tight, with volume concentrations
approaching 70% w/v in sperm
heads, virus capsids and bacterial
nucleoids
1-3
. The packaging
mechanisms adopted may be related
to the natural tendency of semi-
rigid polymers to form liquid
crystalline phases in concentrated
solutions
4-8
. We find that DNA
forms at least three distinct liquid
crystalline phases at concentrations
comparable to those in vivo, with
phase transitions occurring over
relatively narrow ranges of DNA
concentration. A weakly birefrin-
gent, dynamic, ‘precholesteric’
mesophase with microscopic
textures intermediate between those
of a nematic and a true cholesteric
phase forms at the lowest concen-
trations required for phase separa-
tion. At slightly higher DNA
concentrations, a second
mesophase forms which is a
strongly birefringent, well-ordered
cholesteric phase with a concentra-
tion-dependent pitch varying from
2 to 10 µm. At the highest DNA
concentrations, a phase forms
which is two-dimensionally ordered
and resembles smectic phases of
2
Nature
thermotropic liquid crystals
observed with small molecules.
Although specialized proteins
are involved in many DNA packag-
ing processes descriptions of
mechanisms of DNA packaging
must also consider the intrinsic
tendency of the stiff DNA chain to
fold when confined to a small
volume. Stiff nonelectrolyte
polymers form highly ordered,
liquid crystalline phases above a
critical concentration dependent on
the persistence length
4-8
. Formation
of nematic or cholesteric liquid
crystalline phases by semi-rigid
polymers, such as polybenzyl-L-
glutamate, in organic solvents has
been studied in detail
7,8
, and was
predicted theoretically by Onsager
4
,
Flory
5,6
and others. Much less is
known about the behavior of semi-
rigid polyelectrolytes. Because
DNA, fibrous proteins and certain
polysaccharides are polyelectrolytes
involved in numerous supramolecu-
lar associations in vivo, an under-
standing of the phase behavior of
these biopolymers is of fundamental
biological importance.
Ordering of semi-rigid polymers
at high concentrations occurs
spontaneously to minimize the
macromolecular excluded volume
4-8
.
Fig. 2 Multiple liquid crystalline DNA phases observed by polarized light microscopy. A sample of 100-mg ml
-1
DNA in 0.25-M
ammonium acetate was placed on a slider under a partially sealed coverslip so that slow evaporation created a continuous
concentration gradient. Samples were observed through crossed polarizers with a Nikon Optiphot-Pol microscope under 3,200 K
tungsten-halide illumination and photographed on Fujichrome 64-T professional film. a,c, Low magnification views illustrating
sharp transition zones between precholesteric and cholesteric domains (a) and between cholesteric and smectic-like domains (c).
Magnification ~ x 40. b,d-f, Higher magnification views of the three major phases. b, Weakly birefringent, diffuse ring texture
of ‘precholesteric’ phase. d, Highly birefringent, fringe or chevron texture of magnetically aligned cholesteric phase. e, Focal-
conic-fan texture of smectic-like phase. f, ‘Pleated ribbon’ texture of most concentrated smectic-like phase near open end of
coverslip. Magnifications ~ x 100 in b,e,f; x 200 in d.
Semi-rigid polyelectrolyte behavior
is complicated by charge-shielding
requirements. A strong polyelectro-
lyte is surrounded by a counterion
layer, which determines the effec-
tive axial ratio and excluded
volume
9-13
. Because polymer phase
behavior depends on the effective
polymer dimensions, the critical
concentration for DNA ordering is a
sensitive function if ionic strength
and counterion type.
Previous studies demonstrated
that aqueous solutions of persis-
tence length DNA (~500 Å), with
0.3 M NaCl as the supporting
electrolyte, form biphasic liquid
a
b
c
f
d
e
3
Nature
crystalline solutions containing
spherulites at DNA concentrations
(C
D
s) of 130-170 mg DNA ml
-1
at
room temperature, and become
fully liquid crystalline at higher C
D
(refs 14-16). Phase transition
boundaries were in good agreement
with predicted rigid rod behavior
when DNA was treated as a scaled
rod with an effective radius of 21.5
Å at this ionic strength
16
. Here we
report the phase behavior when
DNA solutions over a range of C
D
s
from 100 to 350 mg DNA ml
-1
, as
determined by solid-state
31
P NMR
spectroscopy, polarized light
microscopy and by electron
microscopy.
Solutions showed a uniform
liquid crystalline phase at C
D
s from
170 to 220 mg ml
-1
. A C
D
of 170-mg
ml
-1
corresponds to an effective
DNA volume fraction of 0.76,
assuming 21.5 Å for the effective
DNA radius
16
. Extrapolation to
higher C
D
s yields an apparent
effective volume fraction > 1.0 for
C
D
s exceeding ~230 mg ml
-1
. As
DNA solutions were prepared with
C
D
s over 300-mg ml
-1
, there must be
mechanisms for reducing the
effective DNA volume fraction.
Three potential mechanisms for
reducing the effective DNA volume
are evident: (1) a change to a more
ordered phase, (2) a contraction of
the counterion layer, or (3) a change
in DNA confirmation/hydration. A
phase change above C
D
s of 220-mg
ml
-1
was indicated by a change in
31
P
resonance linewidth and shape
(Fig. 1). At lower C
D
s, in the
biphasic region, the resonance
linewidth increased progressively
with C
D
until the solution became
fully liquid-crystalline, then
dramatically sharpened. Concur-
rently, the lineshape changed from
lorenzian to an asymmetric form
excepted from an aligned sample
with some rotational averaging.
Magnetic alignment of liquid
crystalline DNA phases with the
long molecular axes perpendicular
to the field was reported previ-
ously
16,17
. Magnetic alignment of
Fig. 3 Evolution of the smectic-like phase. A 350-mg
DNA ml
-1
sample in 0.25-M ammonium acetate, when first
applied to a slide and coverslipped exhibited a mottled planar
texture typical of a homoeotropic alignment. Small batonnets
formed within 1-h (a) and enlarged (b), eventually merging to
form a classical focal-conic-fan texture (c). Microscopy and
photography as in Fig. 2. Magnification ~ x 125.
a
the mesophase in a biphasic
solution is prevented because
surface tension restricts the
mesophase to spherulites in which
DNA molecules are oriented
tangentially to the spherical
surface
18
. By analogy, we attribute
lineshape changes at C
D
s slightly
above 220-mg ml
-1
to formation of a
biphasic solution in which two
liquid-crystalline phases coexist.
Magnetic alignment may be
ineffective because of dynamic
exchange of molecules between
phases with different alignment
properties. This conclusion was
enforced by the behavior at yet
higher C
D
s (Fig 1). The resonance
remained broad at C
D
s from 229 to
240 mg ml
-1
, then narrowed again at
C
D
s of 250 and 275-mg ml
-1
,
suggesting formation of a second
uniform, aligned phase. Resonance
broadening indicative of another
phase transition was again observed
at a C
D
of 309-mg ml
-1
.
Interpretations of lineshape
changes in terms of phase transi-
a
c
b
4
Nature
tions were supported by microscope
examination. Liquid-crystalline
phases are traditionally character-
ized by their microscopic textures
as observed through crossed
polarizers
19-21
. Cholesteric liquid
crystals, in which the long axes of
the molecules lie in pseudo-planes
that are slightly twisted with respect
to each other, have periodic varia-
tions in refractive index and fringe
patterns with spacing of P/2, where
P is the cholesteric pitch. Many
smectic liquid crystals show focal-
conic fan textures due to ordering
in a second dimension, and can be
easily distinguished from nematic
or cholesteric forms
20,21
.
Three distinct phases were
observed simultaneously by placing
a 100mg DNA ml
-1
sample on a slide
and partially sealing the coverslip,
allowing slow evaporation to create
a gradient in C
D
across the slide
(Fig. 2 a,c). The low C
D
mesophase
was weakly birefringent and gave
roughly circular patterns of broad
alternating light and dark lines
when viewed through crossed
polarizers (Fig. 2 a,b). Rapid,
small-scale changes in birefrin-
gence due to fluctuations in the
nematic director were noted at high
magnification (x400). The weak
birefringence and director fluctua-
tions imply a dynamic character of
this phase. Thicker specimens of
this phase had an ‘oily streak’
texture and could be magnetically
aligned, but were easily perturbed
when removed from the field.
We tentatively refer to the first
mesophase as a ‘precholesteric’
phase. A weakly birefringent
precholesteric mesophase in
moderately concentrated samples of
high relative molecular mass DNA
was recently reported
19
. The
textures of the latter phase were
attributed to a three-dimensional
Fig. 4 Freeze-fracture-etch electron microscopy illustrating the layered structure of
the smectic-like, high concentration DNA phase. Low magnification view shows
cleavage ‘steps’ (scale bar, 1µm). The direction of shadowing is indicated by the
arrow. DNA molecules appear to lie ‘end-on’ to the viewer and approximately
perpendicular to the layer planes, as expected for a smectic phase. The ‘end-on’
view is readily distinguished from ‘side-on’ views obtained on examination of
cholesteric phases, which tend to shear across cholesteric planes (M.W. Davidson
and R. L. Rill, unpublished observations). Samples were placed on gold grids, quick
frozen between gold planchets using propane jets, fractured at -120ºC, etched at
-100ºC for 2 min, and platinum-carbon shadowed at 45ºC (Balzers BAE 360).
Micrographs of replicas were obtained on a Jeol 100CX in transmission mode at
80 kV acceleration voltage.
helicoidal arrangement of DNA
molecules somewhat analogous to
the molecular arrangement in blue
phases of small molecules. Micro-
scopic textures we observed for the
precholesteric phase of short DNA
were unlike those reported
19
,
suggesting that evolution of DNA
phases may be molecular-length
dependent.
The periodic fringe and ‘oily
streak’ textures of the first
mesophase are reminiscent of a
cholesteric phase, but they differed
significantly from the classical
cholesteric textures of the interme-
diate DNA mesophase, which was
highly birefringent with very
regular, finely spaced fringe
patterns and was extensively
aligned in a magnetic field (Fig.
2d). Freeze-fracture-etch electron
microscopy and optical diffraction
studies have confirmed that this
intermediate phase is, in fact,
cholesteric (M.W. Davidson and
R. L. Rill, manuscript in
preparation).
Smectic phases have two-
dimensional order, with the mol-
ecules arranged in planes at a
defined angle to the preferred
orientation direction of the long
molecular axes. There are two
common ‘natural’ textures of
smectic a and several other smectic
phases-a uniform, homeotropic
texture with molecules oriented
nearly perpendicular to the slide
surface, and the focal-conic fan
texture
20,21
. Focal-conic and
striated fan textures were observed
for the highest concentration
mesophase (s) in controlled drying
experiments (Fig. 2 c,e, and f).
Close examination of a 350 mg
DNA ml
-1
sample confirmed that
DNA forms a mesophase with
molecular layering resembling that
found in smectic phases of small
molecules. Specimens observed
immediately after placement on a
slide usually exhibited a nearly
uniform homeotropic texture.
Batonnets initiated after a few
hours, enlarged, and eventually
merged to yield a classical
5
Nature
focalconic fan texture (Fig. 3).
Examination in the electron micro-
scope of freeze-fracture-etch
replicas confirmed that DNA
molecules were arranged in layers,
and that the long molecular axes
were approximately perpendicular
to the layer planes (Fig. 4). More
definitive data are required to
unambiguously relate this DNA
phase to conventional smectics.
Using light and electron micros-
copy we observed several other
textures in samples of
~300-400 mg ml
-1
(for example,
compare Fig. 2 e,f), suggesting the
existence of other higher-order
phases. As DNA is chiral and packs
hexagonally in crystals
22
, chiral
phases could form which are
analogous to the chiral smectic C
phase, or to more ordered chiral
hexatic phases of small molecules
(for example, chiral smectic G or I
(refs 20,21). These observations
demonstrate that DNA in vitro can
assume multiple-packing arrange-
ments, which can be readily altered
over small ranges of concentration.
We expect that the phase behavior
of DNA in vitro will also be a
sensitive function of the ionic
environment. Significant modula-
tion of DNA packing in vivo may be
accomplished by small changes in
DNA concentration of counterion
atmosphere. Similar principles may
apply to the ordering of other
biological polyelectrolytes.
We thank Dr. Henry Aldrich of
the University of Florida for
assistance with freeze-fracture-etch
methods and Drs. David Van Winkle
and Francoise Livolant for advice
on liquid crystal theory. This work
was supported in part by the NIH.
REFERENCES
1. Du Praw, E.J. DNA and Chromo-
somes (Holt, Rinehart & Winston,
New York, 1970).
2. Earnshaw, W.C. & Casjens, S.R.
Cell 21, 319-331 (1980).
3. Sipski, W. & Wagner, T.E.
Biopolymers 16, 573-582 (1977).
4. Onsager, L. Ann. N.Y. Acad. Sci.
51, 627-659 (1949).
5. Flory, P.J. Proc. R. Soc. A234, 60-
89 (1956).
6. Flory, P.J. in Polymer Liquid
Crystals (eds. Ciferri, A., Krigbaum,
W.R. & Meyer, R.B.) 103-112
(Academic, New York, 1982)
7. Robinson, C. Trans Faraday Soc.
52, 571-592 (1956).
8. Miller, W.A. Rev. Biophys. 29,
519-535 (1979).
9. Manning, G.S.Q. Rev. Biophys.
11, 170-246 (1978).
10. Stigter, D. Biopolymers 16,
1435-1448 (1977).
11. Stigter, D. J. phys. Chem. 82,
1603-1606 (1978).
12. Brian, A.A., Frisch, H.L. &
Lerman, L. S. Biopolymers 20,
1305-1328 (1981).
13. LeBret, M. & Zimm, B. H.
Biopolymers 23, 271-285 & 286-
312 (1984).
14. Rill, R. L., Hilliard, P.R. & Levy,
G.C. J. biol. Chem. 258, 250-256
(1983).
15. Rill, R.L. Proc. natn. Acad. Sci.
U.S.A. 83, 342-346 (1986).
16. Strzelecka, T. E. & Rill, R. L.
J. Am. chem. Soc. 109, 4513-4518
(1987).
17. Brandes, R. & Kearns, E. R.
Biochemistry 25, 5890-5895
(1986).
18. Lerman, L.S. Cold Spring
Harbor Symp. quant. Biol. 38, 59-
74 (1974).
19. Livolant, F. J. Physique 48,
1051-1066 (1987).
20. Demus, D. & Richter, L. Textures
of Liquid Crystals (Verlag Chemie,
New York, 1978).
21.Gray, G.W. & Goodby, J.W.G.
Smectic Liquid Crystals (Heyden,
Philadelphia, 1984).
22. Lerman, L.S. Wilkerson, L.S.,
Venable, J.H. J. molec. Biol. 108,
271-293 (1976).
