- .--
criticali~eviewsin Biochemistryand Molecular Biology. 26(2):151-226 (1991)
Local Supercoil-Stabilized DNA Structures
Max-Planck lnstitut fijr Biophysikalische Chemie, Gbttingen, BRD and Institute of Biophysics,
Czechoslovak Academy of Sciences, 61265 Brno, CSFR
Referee: . " James E. Dahlbwg, Dept. of PhyrlologlcalChemistry, 587 Med. Scl. Bldg., UnlvwsHy of Wlrconaln. 1360;
UnlvsraltyAve., Madison, WI 53706
ABSTRACT:The DNA double helix exhibits local sequence-dependentpolymorphism at the level of the single
base pair and dinucleotide step. Curvature of the DNA molecule occurs in DNA regions with a specific type
of nucleotide sequence periodicities. Negativesupercoiling induces in vitro local nucleotidesequence-dependent
DNA structures such as cmciforms, left-handed DNA, multistranded structures, etc. Techniques based on
chemical probes have been proposed that make it possible to study DNA local structures in cells. Recent results
suggest that the local DNA structures observed in vitro exist in the cell, but their occurrence and structural
details are dependent on the DNA superhelicaldensity in the cell and can be related to some cellular processes.
KEY WORDS:supercoil-stabilized DNA structures, DNA double helix polymorphy, probing of DNA structure,
DNA structure in cells.
I. INTRODUCTION
Until the end of the 1970s, it was generally
accepted that the DNA double helix is very regular
and independent of the nucleotide se-.'
quence.'" This conclusion was based mainly on
data obtained by means of the X-ray fiber diffraction
technique that had been used to study
DNA structure for more than 2 decades. During
the 1960s and 1970s,evidence based chiefly on
the results of empirical techniques gradually
mounted,b10e.g., suggesting that the structure
of the DNA double helix is sequence dependent
and influenced by environmentalconditions.lo In
the early 1970s Bram11.12reached a similar conclusion
based on his studies using X-ray fiber
diffraction. Due to its limited resolution,'.this
technique yields only an averaged DNA conformation;
it cannot detect local variations in the
double helix induced by the particular nucleotide
sequence.I3Using this technique and DNA samples
jith e x e k e s & base composition, however,
Bram12 was able to predict an almost infinite
polymorphy of DNA in the B state. At
about the same time, Pohl and k ~ i n ' ~ . ' ~obtained
circular dichroism (CD) spectra of poly(dGde).poly(dG-dC),
which suggested that this polynucleotide
at high salt concentrations assumes
a structure differing from B-DNA and possibly
left-handed.
The untenability of the single DNA structure
conception became obvious in the mid-1970s.
Based on results obtained with various techniques,
it was suggested that the DNA double
helix is polym~rphic,'~.~~depending on the duplex
nucleotide sequence and its anomalies as
well as on environmentalconditions.1° This conclusion,
however, received little attention at the
time of its publication. :
The situation changed dramatically by the
end of the 1970s, when the first results from
single-crystal X-ray analysis of short deoxyoli-
1040-923819113.50
8 1991 by CRC Press, Inc.
gonucleotides were reported. Unlike &n, crystals
are ordered in three dimensions and can diffract
X-rays at or near atomic resolution, providing
substantially more data. Due to these-factors,
minute details of the DNA double helix
can be observed. The crystal structure of
d(pATAT) with different sugar phosphate conformation
at adenine and thymine residues reported
by Viswarnitra et al.16 in 1978 led to the
proposal of an alternating B-DNA structure for
p o l y ( d ~ - d ~ ) p l ~ ( d ~ - d ~ )with a dinucleotiderepeat
unit.'"18 The left-handed Z-DNA structure
of d(CGCGCG)19and d(CGCG) c r y ~ t a l s ~ ~ , ~ l
solved at high resolution came as a surprise and
immediately transformed DNA structure studies
into a flourishing field. Shortly afterward it was
shown that the structure of a DNA dodecarner
d(CGCGAA'ITCGCG)was its
structure differed, however, from that deduced
by studying DNA fibers. Particularly interesting
was the dependence of local structure on the nucleotide
sequence. At the present time close to
80 deoxyoligonucleotidescontaining four or more
base pairs have been studied by single-crystal Xray
analysis, conclusively demonstrating the nucleotide
sequence-dependent polymorphy of the !,
DNA double helix."
Further aspects of DNA structure polymorphy
were uncovered soon after the discovery of
Z-DNA. These included sequence-directed DNA
curvat~re~~.~'and local DNAstructuresstabilized
by superc~iling.~~-~~In 1980, correlation analysis
of chromatin DNA nucleotide sequence revealed
that certain dinucleotide~,~~e.g., AA or 'IT, occur
at regular intervals correlated with the pitch
of the DNA double helix. The regular in-phase
occurrence of these nucleotides was supposed to
be responsible for the unidirectional curvatureof
DNA in chromatin. A few years later the anomalously
slow electrophoretic mobility of a DNA
fragmentfrom Leishmaniatarantolaekinetoplast
was explained by DNA c~rvature.~',The determination
of the nucleotide sequence of 'this'fragment
revealed runs of adenines and thymines in
phase with turns of the DNA helix.
CruciformsZ8-'I and left-handed DNA
segment^^^.^^ were among the earliest discovered
supercoil-stabilized local DNAstructures. Anecessary
condition for the formation of this type of
structure is a suitable nucleotide sequence and a
superhelix density sufficiently negative to stabilizethe
given structure. In the past decadegreat
progress has been made in the elucidation of the
relations between the DNA supercoiling, local
structures, and their dynamics, interactions, extrusion
kinetics, and other properties (reviewed
in References 31 and 34 to 36). On the other
hand, understanding the biological role of local
structures has lagged considerably.Greater progress
can be expected in this area within the next
few years due to the recent development of research
techniques for the study of DNA structure
in the Thisarticledeals mainly with local
DNAstructuresstabilized by supercoilingin virro
and in the cell; other aspects of DNA structure
polymorphy are briefly summarized.
II. MICROHETEROGENEITYOF THE DNA
DOUBLE HELIX FORMS
Studies of the detailed relationships between
nucleotide sequence and DNA structure became
feasible by the end of the 1970s, when organic
synthesis had been developed to the point that
deoxyoligonucleotidescould be produced in the
purity and quantity necessary for the preparation
of single crystals for X-ray diffraction (and nuclear
magnetic resonance, [NMR]) studies."
Three main families of DNA forms were identified
by crystallographic analysis of deoxyoligonucleotides(for
review see References 13,25,
and 45 to 47): right-handed A- and B-forms and
the left-handed Z-form.
A. A-, B-, and 2-Helices
The A-, B-, and Z-helices have distinctly
different shapes that are due to the specific positioningand
orientation of the bases with respect
to the helix axis. In A-DNA the base pairs are
displaced (0.4 nm)from the helix axis, the major
groove is very deep, and the minor groove is
very shallow. In B-DNA the major and minor
grooves are of similar depths and the helix axis
is close to the base pair center. In 2-DNA the
minor groove is deep and the major groove is
convex. In A- and B-DNA a single nucleotide
can be considered as the repeat unit, while in Z-
DNA the repeat unit is a dinucleotide. In Aduplexes
base pairs are heavily tilted in contrast
to base pairs in B-duplexes, which are almost
perpendicularto the helical axis. Definitionsand
nomenclature of nucleic acid structure parame:
ters were published recently.48
The distinguishing averaged helical parameters
of the DNA forms are given in Table 1.
Many of the structural differences between the
L+ i.
-ik helices arisefrom the puckering of thesugar ring;
-pL C3'-endo is typical for A-DNA, while in 2-DNA
C3'-endo alternates with C2'-endo. In B-DNA
sugar pucker tends to favor the C2'-endo or C1'e
x ~ : ~but the distribution of conformations is
much broader than in A- and Z-DNA. The righthanded
A- and B-forms have the anti glycosidic
bond, whereas in the left-handed 2-helix the orientation
alternates between syn (for purines) and
anti (for pyrimidines). In the latter structure the
orientation around the C4'-C5' bond with respect
to the C3' atom alternates between gauche + and
trans conformations for cytidine and guanosine,
respectively. The alternating features of Z-DNA
result in the zigzag shape of its sugar-phosphate
backbone, from which the name was derived.
Thechangesin the backboneandglycosidic-bond
conformations are accompanied by substantial
variations in the stacking interactions between
successive base pairs in 2-DNA. Methylation or
brornination of cytosines at position 5 (studied
mainly in oligonucleotides with alternatingC-G
sequence) stabilizes 2-DNA. Under certain conditions
even nonalternatingsequences of purines
and pyrimidines can assume the conformation of
2- with thymines in a syn orientati~n.~~.~The
outer surface features of such a 2-helix are different
at the nonalternating sites, but the backbone
is similar to that observed with alternating
sequences.
B. Local DNA Structure and Nucleotlde
Sequence
Average helix parameters for some righthanded
structures are given in Table 2. The significant
variations in some of these global parametersdependenton
nucleotidesequenceresult
in local changes along the DNA double helix.
Such relations have been analyzed in detail by
several authors and reviewed by Shakked and
Rabinovich.13In A- and B-DNA these variations
seem to be determined mainly by the specific
interactions between the stacked base pairs and
also to-someextent bylneighboring bases. In particular,
homopblyme? dinucleotide steps show a
TABLE 1
Comparison of A-, B-, and 2-DNA
A-DNA. B-DNA' B1-DNAb 2-DNA'
Helix sense right-handed right-handed right-handed left-handed
Base pairs per turn
Helix twist (")
Rise per base pair (A)
Helix pitch (A)
Base pair tilt (")
P distance from helix axis (A)
Glycosidicorientation
Sugar conformation
11
32.7
2.9
32
13
9.5
anti
C3'-endo
10
36.0
3.4
34
0
9.3
anti
Wide range
10
34.1, 36.8
3.5, 3.3
34
0
9.1
anti
C2'-endo
12 (6 dimers)
- 10, -50
3.7
45
- 7
6.9, 8.0
anti, syn
C2'-endo, C3'-
end@
' ,
Numerical values for each form were obtained by averaging the global parameters of the
corresponding double-helix fragments.
B1-DNAvalues are for a double helix backbone conformation alternating between conformational
states I and 11.
The two values given correspond to CpG and GpC steps for the twist and P distance values,
to cytosine and guanosine for the others.
TWO valuescorrespond to thetwoconformationalstates.From Kennard,0.and Hunter,W. H.,
O. Rev. Biophys.,22, 3427, 1989. With permission.
TABLE 2
Average Helical Parameters for Selected Right-Handed Structures
A-form
d(GGTATACC)
d(GGGCGCCC)
d(CTCTAGAG)
r(GCG)d(TATACGC)
r(UUAUAUAUAUAUAA)
Fiber A-DNA
9-form
d(CGCGAATrCGCG)
d(CGCGAATT9rCGCG)
Fiber B-DNA
Rise per Groove width
Helix base pair Base pair Propeller (A) Displacement
twist (") . (A) tilt (") twist (") Minor Major Da (A)
BrC = 5-bronecytosimo.
Adapted from Kennard, 0. and Hunter, W. N., Q. Rev. Biophys., 22, 327,1989. With permission.
wide spectrum of stacking characteristics that are
markedly neighbor dependent. On the other hand,
pyrimidine-purinesteps in A-DNA (especially the
C-G steps) often display a low twist and high
slide that are only slightly dependent on lieighboring
steps. Ln 2-DNA the shape of the helix
surface changes significantly due to deviations in
the regular alteration of the purine-pyrimidine
sequence, while the sugar-phosphate backbone
does not change. The effect of the nucleotide
sequence on the fine geometricalfeatures of each
DNA form has been clearly demonstratedbut not
fully elucidated. The emerging rules, however
should be considered as tentativesince they were
based on a relatively small number of examples.
The well-known "Calladine's rules"51 are now
perceived to be incomplete and to neglect important
factors other than the steric clash of purine
r i n g ~ . ~
C. DNA Hydration
Information about the organization bf water molecules
in DNA forms has recently been gained
(reviewedin References 25, 52, and 53) from Xray
diffraction analysis of crystals. Distinct hydration
patterns were observed in the major and
minor grooves and around the sugar-phosphate
backbone. It was proposed that in DNA with a
mixed nucleotidesequencehydrationof the backbone
is related to global conformati~n.~~In Aand
2-DNA a chain of water moleculescan bridge
the phosphate oxygens along the backbone. There
are more water molecules around each phosphate
group in the B-DNA, but almost no water bridges
.between the phosphate oxygens, as the distance
betweehphosphate oxygens in this DNA form
are too great to be linked with a single water
molecule. It appears that specific nucleotide sequences
that create local changes in the DNA
double helix may also affect the backbone hydration
pattern.54-55Even greater dependence of
the hydration patternson nucleotidesequencehas
been found in the DNA grooves. In A-DNA specific
hydration patterns occur in the major
grooves.25A string of well-ordered water molecules
hydrogen bonded to oxygen and nitrogen
atoms in the minor groove has been found in the
central AAlT sequence of the B-DNA dodecamer
d(CGCGAATTCGCG).56.57 This specific
hydration of B-DNA, the "spine of hydration",
significantly contributes to DNA stability. Studies
of further B-DNA helices (Table 3) revealed
two ribbons of water molecules along the walls
of wide regions of the minor groove, while narrow
regions of the minor groove contained an
ordered zigzag spine of hydration.* It appears
that the interdependence between nucleic acid
structure and the solvent represents one of the
TABLE 3
Summary Comparison of Properties of &DNA Helices
Nucleotlde I.D. Resolutlon Minor groove Mean propeller Minor groove Helix bend at1
sequence c o d e (A) wldthD
twist" hydratlond
(if present)
W N W L H L b S b
b u b
b u b
b u b
Bent, CGCGt
AAlTCGCG
Bent, CGCI
AAAllTGCG
Bent, CGCt
AAAAAAGCG
Bent, CGC/
AAAAATGCG
Straight
Bent, CGCt
ATATATGCG
Bent, CGCGt
ATATCGCG
Straight
Straight
Straight
Straight
Straight
W N W L H L
CGCAAAAAAGCG W N W L H L
CGCAAAAATGCG W N W L H L
CGCGAAlTBrCGCG
CGCATCTCTGCG
W N W
W N W
L H L
1
b S b
b u b
CGCGATATCGCG W N W L I L
GpsCGpsCGpsC
CCAAGAlTGG
CCAACGlTGG
CGATCGATCG
CCAGGCCTGG
W
lwwwl
WWNWW
WNWNW
IWlWl
L
L H L
L I L
L I L
L (1) L
R
R
R S R
S R S
R S R
Identification code usually from initials of first author: HD = Horace Drew; MC = Miquel Coll; HN = Hillary Nelson; AD =
Anna DiGabrieli;MK = Mary Kopka; CY = Chun Yoon; ZS = Zippora Shakked (personal communication);WC = William
Cruse; GA,CG = Gilbert Prive; KK = Kazunori Yanagi and Kasimietz Gtzeskowiak; UH = Udo Heinemann.
W = wide minor groove (>I2A), I = intermediate minor groove, N = narrow minor groove (el0A).
= H = high propeller twist (>15"), 1 = intermediatepropeller twist, L = low propeller twist (<I 0").
* S = central spine of hydration; R = two ribbons along groove walls; d = hydration irregular or disordered; b = groove
blocked, no information; u = hydration state uncertain because of lack of resolution. From Dickerson, R. E. in Structure
and Methods, Vol. 3, DNA andRNA, Sarma, ,R. H. and Sarma, M:H., Ed~.,~AdeninePress, Schenectady, N.Y., 1990,1.
With permission. i ;
bases for DNA double helix polymorphy (reviewed
in References 25, 52, and 53).
major groove width, rise per base pair, and base
pair inclination change in an almost continuous
fashion upon going from A- to B-DNA. A- and
B-DNA can be distinguished, however, by lateral
displacementof the base pairs from the helix axis
(Table 2) that changes discontinuously between
the two helical families. It has been shown that
in the crystal structure of d(GGBrUABrUACC)
A and B conformations can coexist.j9Coexistence
of these two forms was detected also in
d(GGGGG?-I-I-I-r)-d(AAAAACCCCC) in 6 M
NaCl by Raman spectroscopy showing an A-B
junction occumng over a small span of bases.60
This may imply that adjacent segments of DNA
might adopt different global conformations in
vivo, for example, as a result of a specific DNA
protein interaction in the absence of drastic environmental
changes.
D. Continuum of Right-Handed DNA
Conformation
The original concept of unique global conformations
of right-handed DNA based on X-ray
fiber diffraction analysis is undergoing significant
change. Instead of discontinuous states of
A-, B-, and C-forms that are stable only in conditions
that differ significantly from one another,
a continuum of right-handed DNA conformation
is being con~idered.~.~~A recent comparison of
the published A- and B-crystal structures
revealeds8 that certain structural features such as
E. DNA Structure, Crystal Packing, and
Environmental Conditions .
In spite of the great progress made in the
crystallographic analysis of DNA in the last decade,
some basic problems still remain to be clarified,
including questionsof the influenceof molecular
packing in the crystal and the effect of
temperature, hydration,etc. Quite recently it was
revealed that the DNA molecular conformation
and molecular packing are closely rela~ed.~'The
global DNA structure of the sameoligonucleotide
may vary substantially between different crystalline
states; however, it also appears that the
great sensitivity of DNA structuretocrystalpacking
might be limited to only short A-DNA fragments
(up to about 8 Furthermore, the
structure of a B-DNA decamer can be nearly the
same in two very different crystal lattices.63
Local DNA structure in the given crystal environment
is determined not only by nucleotide
sequence but also by hydration and temperature
e f f e ~ t s . ~ . ~ ~An interestingquestion is the relation
between the DNA structure in crystals and in
solution. GC-rich oligonucleotides often crystallize
as A-DNA,58but their solution structure as
determined by Raman or NMR s p e c t r ~ s c o p y ~ ~ ~ ~ ~
at moderate ionic strengths is predominantly B- \
DNA. A possible explanation may be seen in the
presence of both forms in solution (a smaller
amount of A-form being undetectable due to the
limited sensitivity of spectroscopic techniques)
combined with preferred oligonucleotide crystallization
as A-DNA (which would represent a
more favorable way of crystal packing than the
B - f ~ r m ) . ~ ~Dependenceof the DNAcrystal structure
on t e m p e r a t ~ r e ~ . ~ ~ . ~ . ~ 'and other environmental
conditions as well as partial opening of
the base pairsu in B-DNA under conditions far
below melting might be related to premelting
changes in DNA conformation observed in
solution.lo
The helical axis of a DNA segment can be
unidirectionally bent if tracts of adenine residues
occur at regular intervals (around 10.5 bp) of the
helical repeat (reviewed in References68 to 72).
Bending of DNA can also be induced by an external
force; protein-induced bending has been
reviewed r e c e n ~ l y . ~ ~ . ~ ~Compared with normal
DNA, intrinsically curved DNA has a decreased
electrophoretic mobility;69.71.72-75.76this property
has proven to be a useful experimental criterion,
but it should not be taken as an absolute indication
of DNA curvature. Otherfactorscan result
in altered DNA mobility and under some conditions
curved DNA molecules may display normal
electrophoretic mobility.72 DNA curvature
can also be visualized in the electron micro-
scope.70.77,78
A. Nucleotide Sequence Requirements
and Phasing of the Sequence Motif
From a detailed analysis of the sequence requirements
for DNA bending a general picture
may be drawn:
1. An intact tract of four to six adenine residues
is importantfor the phenomenon, with
six A's supplying the maximum effect.71
2. Interrupting the sequence with G, C, or T
strongly decreases the effect (T is the least
effective);
3. Flanking of the A-tract by T on the 3'-side
and by C on the 5'-side results in the greatest
degree of curvature.
4. Adenine can be replaced by inosine in the
A - t r a ~ t ~ ~ . ~ ~(curvature remained detectable
for a small number of I in sequences such
as AlAIA and AAIAA; it disappeared,however,
in oligomers composed only of 1-C
pairs).
5. Deletion of the 2-amino group from the minor
groove of the B-DNA helix increases
DNA curvature, while displacement of &I
amino group from the major groove to the
minor one may eliminate curvat~re.~'
6 . Methylation of adenine can increase DNA
bending.82Pyrimidine 5-methyl groups influence
the magnitude of the DNA curvat
~ r e ; ~ ~ . ~ ~such effects are position
dependent.83
Proper phasing of the sequence motif is of
crucial importance. Ifthe motif was repeatedevery
9 or 12 bp, DNA bending was significantly reduced,
whereas repetition every 1.5 helical turns
resulted in the disappearance of electrophoretically
detectable bending.8s.86The possibiiity that
the A-tract might provide a flexible universal
hinge for bending was excluded by the latter experiment.
DNA curvature is influenced by environmental
conditions. Increasing ~ernperature~~
or NaCl con~entration~~.~~decreases DNA bending,
while addition of divalent ions increases
bending in some case^.^^.^^
B. Strclctural Models
Intrinsic curvature represents an interesting
aspect of the sequence-dependentpolymorphy of
the DNA double helix. Attempts have been made
toexplain this phenomenonand structural models
proposed. "Wedge models",89 based on Trifonov
and Sussman'sZ6suggestion that the angle
between base planes (the wedge angle) is sequence
dependent, emphasize the smooth deformation
of DNA by a series of small tilt and roll
components between adjacent base pair planes.
Bolshoy et aLgOrecently showed that in addition
to AAw'IT, other base pairstacks, namely, A m ,
CG-CG, and GA.TC, have large wedge angles
that induce appreciablecurvature into DNA molecules
as detected by their anomalous gel mobilities.
They considered the possibility that formation
of transient kinks (as a result of forces
applied to DNA during gel electrophoresis), especially
in CAsTG and ACsGT stacks, may influence
the anomalous mobilitiesof curved DNAs.
In the "junction m~del"~l-~la deflection of the
helix axis is expected at the junction between Band
non-B- (A-tract) DNA. The wedge model is
more general,71as it can include some aspects of
the junction
C. Structure of A-Tracts
Despite numerous studies (reviewed in References
71 and 72), uncertainty remains regarding
the structure of A-tracts. The original suggestion
that poly(dA).poly(dT) adopted a heteronomous
structure
g2
was abandoned after further X-ray fiber
diffraction studies
g3
and reexamination of the
original data.* It was shown that the results of the
X-ray fiber dif3action studies are most consistent
with the B-type helix with a distinctly narrow
minor groove, large positive propeller twist, and
negative tilt of the base pairs (B1-DNA).According
toenergy cal~ulation,~~stackingenergies
of AT pairs in poly(dA)-poly(dT)are not optimal,
but the lost stacking energy seems to be compensated
for by a formation of a water spine in
a minor groove;such a hydration network cannot
form in GC-rich sequences where the amino group
of guanine intrudes into the minor groove.
Single-crystal X-ray analysis of several Atract-containing
DNA duplexes confirmed the
presence of the narrow minor groove and a markedly
pronounced propeller twist of the AT
pair^.^^-^^ In addition, a changed Watson-Crick
base pairing was observed in which the C6 amino
group of adenine formed an additional hydrogen
bond with the 0,of the 3'-neighboring thymidine.
It has been suggested72that the presence of
bifurcated hydrogen bonds is not required for
DNA curvature since AIAIA sequences display
a significant curvature but are not able to form
bifurcated hydrogen bonds.79The single-crystal
X-ray measurements produced important information
regarding short-rangeconformationof the
A-tractbut do not provide unambiguousevidence
for the structural basisof curvature.Theobserved
junctional diitortiohs may be a result of crystal
lattice constraint. It appears that the A-tract curvature
in crystals is due to the intramolecular
packing constraints of the crystal lattice and that
the structureof this tract is dependent on its length.
Lengthening of an A-tract resulted in a cooperative
switch from B-DNA structure to one more
characteristic of poly(dA)*poly(dT);" in longer
A-tracts the minorgroove narrowed over the first
three base steps and then remained almost constant.
Measurements of imino proton exchange
rates showed anomalously long lifetimes in A4and
A S - t r a c t ~ . ~ ~Thymidine protons in shorter
tracts tended to have shifts similar to those of B-
DNA.
Using CD and UV absorption spectroscopy,
Herrera and Chaireslolprovided evidence of temperature
or dimethylsulfoxide-induced premelting
conformational transition of poly(dA).poly(dT).
Their results suggested the
existence of at least two species at equilibrium
underconditions far below the melting transition.
A distinct premelting transition in the duplex
GA,T,C was observed by Park and Bresla~er'~
by means of differential scanning 'calorimetry.
These results are consistent with the hydration
model proposed by Chuprina." Disruption of the
spine of water in the minor groove induced by
the temperature or dimethylsulfoxide would result
in loss of curvature and normalizationof the
electrophoretic mobility actually observed at elevated
temperatures.69
D:' Biological Role
Intrinsic DNA curvature has attracted great
attention in recent years, but the question of its
biological role has not been solved. Numerous
papers have been published (reviewed in Reference
72) reporting loci of curvature in a wide
variety of organisms, in many cases in regions
of functional importance. Poly(dA).poly(dT) and
DNA containing A,,-tracts from Leishmania tarantolae
were shown to be imrnunogeni~'~~and
specifically detectable by antibodies. Yet, is a
curvature per se of biological importance, or is
it the A-tract sequence only that has an important
function? Recent results obtained by Linial and
S h l ~ m a i ~ ~ ~ - ' ~ 'suggest that the endonuclease from
C.fasciculata binds in a specific manner to curved
DNA, recognizingspecific local variationsin helix
trajectory. It has been shown that a number
of curvature loci appear near prokaryotic (and
Simian virus 40, SV40)origins of replication and
that the degree of curvature of such loci is increased
due to binding of specific replication proteins,
suggesting a possible relation between curvature
and repli~ation.~Transcription activity can
be significantly influenced by the presence of a
curved DNA segment in the vicinity of a promoter.lOB.'@'Binding
of the catabolite activator
protein to upstream sites in many E. coli promoters
induces DNA bending and transcription
activation. Introduction of intrinsically curved
DNA upstream from the promoter resulted in
transcription activation in the absence 'of cata- '
bolite activator protein binding,'@'suggesting that
DNA curvature plays an important role in the
transcription activation. Involvement of DNA
curvature in DNA recombination was also demonstrated
recently.'lo.' I '
Considering the biological relevance of
curved DNA, the properties of negatively supercoiled
DNA-containing curvature-inducing sequences
are of particular interest; however, little
is known about the relations between supercoiling
and DNA curvature. It has been shown that
the presence of curved segments can influence
writhing of the supercoiled DNA." Minicircle
kinetoplast DNA-containing curvature-inducing
sequences were linearized by single-strand selective
mung bean nuclease in the presence of
50% f~rmamide."~Supercoiled plasrnid pK5/
6T217 was site-specifically modified under
physiological conditions with osmium tetroxide,
pyridineH3within the curvature-inducing sequence
dATATAIITITITAGAGATTTTT. Such
site-specific modification was detected neither in
linearized pK5/6T217 nor in supercoiled and linearized
plasmids containing a slightly less curved
dGACAAAACTC sequence. The site-specific
modification pK516T217 with the single-strand
selective chemical probe suggests structural distortion
in the curvature-inducing sequence resulting
from the superhelical stress. It is, however,
not clear whether this distortion is related
to DNA curvature or only to the specific AT-rich
.sequence (see Section VII.D.2).P
i i
IV. DNA SUPERCOILING
Vinograd and co-workersH4
reported the discovery
of the twisted circular form of polyoma
viral DNA 25 years ago. Their results wereclosely
connected to the earlier work of WeilH5and
Leb~witz,"~who had demonstrated some unusual
properties of polyoma DNA, as well -asto
the findings of Vogt and Dulbecco,"' who had
shown that conformation of this DNA can be
changed by a single DNaseI cleavageevent. Weil
and Vin~grad,"~in collaboration with Stoeckenius,
visualized three forms (linear, open circular,
and twisted) of polyoma DNA in the electron
microscope. By means of a series of
experiments that 'included measurements of pHinduced
changes in DNA sedimentation velocity,
Vinogradand co-workers114explained the twisted
appearance of polyoma DNA by the formation
of superhelical turns in the circular molecule. It
soon became apparent that circular DNAs from
many sources were supercoiled and that even linear
DNA could supercoil inside the cell if constrained
into topologically distinct domain^^^^.^^^
(reviewed in References31,45,and 121to 125).
A. Decomposition of Linking Number to
~wistingand Writhing
In linear DNA, the ends are free and the
moleculecaneasily accommodatechangesin helical
pitch. If, however, the two strands become
linked to form a loop or a covalently closed circular
molecule, the number of times the strands
wind about each other (the linking number, Lk)
cannot by changed. Lk is an integer that may be
de~ornposeď~~-~~~as
where Tw (twisting) is the number of times one
strand rotates around the DNA duplex axis, which
corresponds to about 10.5 pb per unit twist for
B-DNA in solution, and Wr (writhing), a measure
of the global tertiary structure of the mol-
'
ecule, describes the deformation of the duplex
axis. If the DNA duplex were able to under- or
overwind easily, Lk would always be equal to
Tw (as is the case with a relaxed DNA) and no
supercoiling would arise. DNA tends, however,
to retain its B-form, thus limiting Tw to a narrow
range, and changes in Lk can be compensated
for by changes in superhelicity and by local
changes in DNA secondary structure and formation
or absorptionof somespecific local structures,
as explicated in the following section.
Closed circular DNA molecules can be characterized
by the linking difference ALk, the difference
between Lk and Lk,
,. .
ALk = Lk - Lk, ' . ' (2)
Lk,,is the linking number of a relaxed reference
DNA (lacking superhelicalturns) under the given
conditions; ALk is often referred to as the number
of superhelical turns. Equation 2 can be transformed
into
where a is the specific linking difference (independent
of the number of base pairs per DNA
molecule), frequently termed superhelix density.
Isomeric DNA forms that result from topological
differences are called topoisomers. Individual
topoisomers of plasmid and viral DNAs
can be resolved by one- or- two 2-D gel electrophorese~~~.'"(due
to the different shapes of each
topoisomer molecule)-(see Section V.A). Each
topoisomer can also be characterized by a difference
in free energy arising from the deformations
required to accommodate the linking difference.
The free energy (AG,) is related quadratically to
A Lic
K.RT 2,
AG, = -
N
ALk
where K is the proportionality constant equal to
1050 for molecules larger than about 2 kb, R is
the gas constant, T is the absolute temperature,
and N is the number of base pairs per molecule.
In naturally supercoiled DNAs the freeenergy is
appreciable. At = -0.05, plasmid pBR322
has a free energy of about 100 kcallrn01.~'In
recent years the mathematical analysis of DNA
supercoiling has expanded ~ignificantly.'~~.'~~-l~
It has been shown that the division of the linking
number into three rather than two components is
more convenient, as this helps to clarify relationships
between experimentally measurable
quantities and topological and geometric DNA
properties. This new subdivision has been recently
broadly developed;'28its detailed discussion
is beyond the scope of this article.
B. Negatively and Positively Supercoiled
DNA
Equation 2 wasapplied to numerouspurposes
that included, in addition to the determination of
the extent of supercoiling, determination of the
free energy associated with DNA s~percoi1ing.l~~
determination of the average duplex rotation angle
in DNA and its dependence on temperature and
salt concentrati~n,~~~.'~~analysis of the ex-
tent of DNA winding in nucleosomes, 141.142
screening of DNA binding modes in various
drugs,143 etc. In most of the naturally supercoiled
DNAs Lk < Lk,such DNAs are called negatively
supercoiled. They are underwound and the
form of theirsuperhelix is right-handed. A typical
superhelix density of isolated plasmid or viral
circular DNAs is around -0.06. Positively supercoiled
(overwound, left-handed superhelical
turns) DNA with Lk > Lk,was recently found
in archaebacteria Sulfolobus acidocaldarius.lM
In E. coli and other prokaryotes, positively supercoiled
DNA can be generated during transcription
in front of the transcription ensemble.145
C. Physical Properties of Supercoiled
DNA Molecules
Propertiesof supercoiledDNA and DNA interactions
with single-strand binding proteins as well
as with other proteins were studied by means of
physical and physicochemical technique^.^^^-^^^
The recent results of Raman spectroscopy measurementsof
supercoiled and nicked ColEl DNA
suggest that accommodation of supercoiling
occurs mainly in AT base pairs and backbone
moieties.l6lRemelting effects might accountfor
these changes, including a slight change in the
band known to be responsiblefor base pair breakage.
The major part of the alteration of the backbone
geometry takes place in the C-0 linkage
between the C5' and adjacent phosphate group.
Raman spectra obtained with supercoiled
pBR322lS9and its derivative ~ F b 1 0 0 ~ ~did not
completely agree with those of ColEl DNA; the
reasons for this disagreement are not known. It
is tempting to suggest that the presence of ATrich
C-inducing sequences3
' in ColEl (See Section
V1.A) might be connected with the observed
Raman spectra of this DNA. Positively supercoiled
pBR322 induced a negative contribution
to CD of the main bands at 270 and 187 nm and
showed a higher electrophoretic mobility when
compared with the negatively supercoiled topisomeric
sample (o= 0.07 and -0.07, respectively).16-'
As expected, cruciform structure did
not extrude in the positively supercoiled DNA.
D. Shape of Supercoiled DNA Molecules
Negatively supercoiled DNA can exist in two
basic forms: solenoidal (toroidal) and plectonemic
(interwound). The latter form has been
assumed to be the most probable form of negatively
supercoiled DNA free in solution. Solenoidal
supercoiling can be exemplified by the
wrapping of DNA along histones in nucleosomes.
There are some important differences between
these two forms. The solenoidal form is
left-handed, while the plectonemic (interwound)
is right-handed. Plectonemic supercoiled DNA
can be naturally branched, bringing together nucleotides
that are distant from one another in the
primary structure; this is not the case with solenoidal
supercoils. It is probable that these two
forms are in a dynamic equilibrium in eukaryotic
cells,. Not all protein-bound supercoils must be
necessarily solenoidal, as the supercoil form is
dictated by the winding surface; well-characterized
plectonemic DNNprotein complex can be
represented by the synaptic intermediate of the
Tn3 resolvase.128*146
Recent computer statistical-mechanical simulations
of moderately and highly supercoiled
DNA molecules (treating supercoiled DNA withinn
the wormlike model with excluded volume)
showed irregularly shaped molecules with characteristics
of branched interwound helices at
higher superhelix densities. la The calculations
showed that the quadratic dependence of the superhelicalfree
energy on Lk (Equation 4) is valid
for a variety of conditions but is not universal.
Significant deviations can be expected at high
superhelix density under ionic conditions where
effective diameter of DNA is small.
Experimental data concerning the shape of
supercoiled DNA in solution are not free of ambiguity.
The results of small-angle X-ray scattering
are consistent with the toroidal shape but
not with an interwound shape,149.150 while dynamic
light-~catteringl~~-l~~and neutron diffraction
in liquid crystalline solution158favor the interwound
superhelical structure. Most electron
microscopy ~ o r k ~ ~ ~ * ~ ~ ~displayed images consistent
with the interwound model, but changes
in the DNA shape due to the DNA absorption
and drying on the supporting film were not excluded.
Quite recent observations by"Adrian et
al.164made by cryoelectron microscopy of vitrified
specimens demonstrated the interwound form
of the protein-free negatively supercoiled DNA.
These authors also showed that the shape of the
supercoiled DNA molecules was strongly affected
by Mg2
+ ions; in 10 mMTris the diameter
of the pUC18 DNA molecules was about 12 nm,
while it was reduced to 4 nm in 10 mM MgCI,.
Approximatevalues of the partition of the linking
difference (ALk) between the changes of writhe
(AWr) and twist (ATw) were calculated from a
single projection of pUC18 topoisomer with a
ALk = -12 (in Tris solution with no added
MgCI,). The partition between ATw and AWr
was estimated to be between 1:3 and 1:4. These
are the frrst data of ATw and AWr partitions
based on direct measurements.
E. Topoisomerases
Supercoiling is controlled by enzymes called
topoisomerases (reviewed in References 166 to
168) that may be divided into two classes. The
type I enzymes do not require an external source
of energy; the transient intemption of a DNA
phosphodiestericbond is accompanied by the formation
of an enzyme-DNAcovalent intermediate
that conserves the free energy of the bond. This
energy is then used for resealing the broken strand
after its rotation around the unbroken strand. Type
IIenzymes transiently break both strands, forming
two covalent bonds between the DNA duplex
and the enzyme molecule. The latter type of enzymes
usually require ATP energy. Besides their
basic ability to modify DNA linking number,
topoisomerases can perform a number of topological
reactions involving both double- and single-stranded
DNAs, folding them into various
kinds of topological knots and catenates.
V. METHODS OF ANALYSIS OF LOCAL
DNA STRUCTURES
In natural supercoiled DNA molecules such
as plasrnids,local structures representonly a very ,
small part of the whole molecule, usually around.
1%. Application of conventional physical techniques
to the study of these molecules is thus
very limited. In the last decade new methods
were developed that have made it possible not
only to detect a local structure in the DNA molecule,
but also to provide its exact location and
information about its chemical reactivity at single-nucleotide
resolution. In the early
~ t u d i e s ~ ~ * ~ - ' ~ . "analysis of DNA electrophoretic
mobility, antibodies, and enzymatic probes represented
the main tools in local DNA structure
research. In the past 5 to 7 years chemical probes
have been increasingly applied, partially replacing
the enzymatic probes.
A. Analysis of DNA Electrophoretic
Mobility
Individual topoisomers can be resolved on
agarose gels (see Section IV). It was shown169
that the electrophoretic mobility of circular DNA
molecules containing long, inverted repeats increased
regularly with supercoiling to a certain
threshold level and then collapsed back.This collapse
corresponded to the extrusion of the cruciform
structure, presumably reflecting the de.crease
in writhing that accompanies the extrusion.
In this way gel electrophoresis can yield information
about cruciform extrusion and other transitions
in supercoiled DNA molecules. In 2-D
gel e l e c t r o p h o r e ~ i s ~ ~ ~ ~ ~ ~ - ~ ~ ~a family of topoisomers
in which ALk ranges from about zero to
relatively large negative values is prepared first.
Then electrophoresis is performed in the frrst dimension,
followed by electrophoresis in the direction
perpendicular to that of the fxst dimension.
The electrophoresis in the second dimension
is carried out in the presence of an intercalator, chloroquine,
which reduces the twisting number
Tw in all topoisomers. This decreases the supercoil-induced
stress, so that any DNA segment
that might have turned into a cruciformor another
unusual structure reverts to its original B-form
DNA. Thus, in the second dimension the mobility
is governed only by the Lk of the topoisomers.
As a result of this procedure an arc of
topoisomers is formed in which a discontinuity
is indicative of the local structural transition. From
the results of 2-D gel electrophoresis, the superhelix
density necessary for the structural transition
as well as the change in Tw induced by this
transition can be calculated. In an experiment
performed by Wang et al.,17' for example, due
to the flipping of a (dG-dC),, segment from a
10.5-fold helix into a left-handed 12-fold 2-helix,
the observed ATw was equal to 6, conesponding
well to the expected decrease by (321
10.5) + (32112) or 5.7. In addition, coupling
between the free energy changes and local structural
transitions in supercoiled DNA molecules
can be ~tudied~'-'~'by 2-D gel electrophoresis,
which" has been applied to studies of crucif
o r m ~ , ' ~ ~ - ' ~ ~Z-DNA,171.174and triplexes175in supercoiled
plasmids. 2-D gel electrophoresis performed
at various first dimension temperatures
provided a sensitive assay for the detection of
local structural transitions in closed duplex
DNAs. '"
The possibility of applying electric fields in
two orientations was utilized in a different way
in pulsed field eletrophoresis, developedover the
past several years for the resolution of very
large DNA molecules (ranging from 20 to 2000
kb).17"178This technique makes possible the separation
of intact chromosomal DNA molecules
from lower eukaryotes as well as large human
chromosomal DNA fragments (References 179
to 181 and references therein) and can also be -
'
applied to studies of closed circular DNA molecules
and their
Another method introduced recently is denaturing
gradient gel electrophoresis,18'-IE6 using
solvents184(urea, formamide), temperat~re,'~~.'~~
or their combina~ion'~~as denaturing agents. The
latter technique was applied to the study of early
melting of supercoiled DNAs.lS8Increased application
of denaturing gradient gel electrophoresis
can be expected in the near future, as this
technique has many assets, including the possibility
of studying DNA structural transitions, even
in partially purified samples and in cell extracts
containing small amounts of DNA.
The procedure based on the difference in
electrophoretic mobility between a specific:DNAprotein
complex and free DNA is widely known
asgel retardationanalysis.189.190In the lastdecade
this technique has been increasingly applied in
DNA-protein interaction studies, including interactions
with antib~dies,'~~-'~"RNA polymera
~ e , ' ~ ~etc. Practical protocols for gel retardation
experiment^'^^ and reviews of recent applic
a t i o n ~ ' ~ ~and factors affecting the lifetime and
mobilities of protein-DNA c~mplexesl~~were
published within the last 2 years. The gel retardation
technique has been widely used in studies
of DNA curvature (see Section 111). The combination
of curved stretches in phase with noncurved
segments makes it possibleto detect small
structural changes involving changes in the DNA
helix axis. 198
B. Antibodies Recognizing Local DNA
Structures
Nucleic acid immunochemistry providesvaluable
reagents for studies of complex biological
materials and of purified nucleic acids (reviewed
in References199 to 201). Most abundant nucleic
acids are only weakly immunogenic, whereas
chemically damaged nucleic acids and less usual
DNA structures show greater effectivity in inducing
antibody formation. The low irnrnunogenicity
of B-DNA may be due to its rapid degradation
by serum nu~leases.'~~On the other hand,
antibodies reacting with B-DNA were found in
sera of patients with systemic lupus erythemat
o s ~ s . ~ " ~Monoclonal antibodies have been isolated
from hybridomas derived from mice with
a disease similar to human lupus erythematosus.
Some of them showed an ability to recognize BDNAs
with different nucleotide sequences.
Local DNA denaturation can be recognized
by antibodies induced by insoluble complexes of
denatured DNA with methylated bovine serum
albumin or by conjugates of base nucleotides or
nucleotides with protein^.^^^.^^^ The most prominent
subject of DNA structure irnrnunochemical
studies has been left-handed 2-DNA (reviewed
in References 34, 199 to 201, and 205). -Antibodies
have been induced mainly by brominated
or methylated poly(dG-dC).poly(dG-dC) in 2form.
Monoclonal antibodies recognizing various
nucleotide sequences or a single sequence have
been generated.
Antibodies to triplex structures have been induced
by triple-strandedpolyribonucleotides and
mixed polyribo- and polydeoxyribonucleotides.*05Lee
et al.206recently generated a monoclonal
antibody against poly(dT-d5"C).poly(dGdA)-poly(dT-dSmC)capable
of forming a triplex
at neutral pHsZo7By use of this antibody, the
presence of triplex structures in supercoiled
plasmids206.208at pH 5 as well as in cells and
was demonstrated. Triplex
structures were also sensitively detected by immun~blotting.~~"Monoclonal
antibodies against
cruciform structures have also' become
available.210
Antibodies represent a very useful tool in
local DNA structure research due to their.high
specificity and sensitivity and applicability in
complex biological media. On the other hand, at
higher concentrationsof antibodies, the equilibrium
between a B-form and a specific local DNA
structure recognized by the antibody (e.g., ZDNA)can
bestronglyshifted in favorof the latter
structure. In cytological studies antibodies are
applied after futation, which may significantly
influence the presence of the DNA structure detectedby
the antibody. These factsshould be kept
in mind when using DNA structure-specificantibodies
to report the presence of local DNA
structures in various biological materials (see
Section VII).
C. Enzymatic Probes
The application of enzymes for probing local
DNA structures is based on the ability of some
nucleases to recognize and cleave DNA at sites
with a more or less single-stranded character
without cleaving the intact B-DNA. Properties
of some single-strand selective nucleases were
previously summarized.'O Among these enzymes,
nuclease S 1 has been most frequently
used for local DNA structure studies.211.212This
enzyme cleaves DNA (at acid pH in the presence
of Zn2
+ ions) in the cruciform loop^,^^.^^ at the
B-Zjunctions,213at triplex sites,175etc. It appears
that the enzyme recognizes distortions in the
sugar-phosphate b a c k b ~ n e , ~ ~ ~ . ~ ~ ~but the exact
structure of the sites is not known. It has been
shown that nucleaseS1does not cleaveat asingle
base mismatch, but two mismatchesaresufficient
for c l e a ~ a g e . ~ ~ ~ . ~ ~ 'Mung bean nuclease displayed
properties similar to those of S1 n u c l e a ~ e ; ~ ' ~ . ~ ~ ~
however, it has been used in local DNA~trycture
studies to a lesser extent.22G223In contrast to these
two enzymes, endonuclease from Neurospora
~ r a ~ ~ a ~ ~ ~ * ~ ~and PI n u ~ l e a s e ~ ~ ~can be applied at
neutral pH. Interest has increased in the latter
enzyme, which is commercially available.226228
Enzymesdo not seem to be ideal DNA probes
for many reasons, includingthe limitation of their
application to the conditions of optimum enzyme
activity, their mostly unknown mechanism of action,
the possibility of induction of secondary
structural changes in DNA, etc. For these reasons,
attemptshave been made to findother DNA
probes.
D. Chemical Probes of the DNA
Structure
A common feature of the local DNA structures
studied to date is the presenceof bases that,
compared with the bases in the B-form, have
better accessibility for interaction with the environment,
e-g., bases in the cruciform loop, at
the B-Z junction, etc. One may thus expect that
these bases will display enhanced reactivity toward
some chemicals. In the beginning of the
1980s229.230we looked for a chemical probe of
the DNA structurethat could (1) bind specifically
to single- and distorted double-stranded DNA regions
under nearly physiological conditions, (2)
form a covalent bond sufficiently stable even after
the removal of the unreacted agent, (3) form a
DNA adduct that,iseasily detectable even if only
a very small friction of bases in the DNA molecule
undergo the reaction, and (4) be potentially
applicable for studies of DNA structure in situ.
We have shown that osmium tetroxide, pyridine
reagent (Os,py) fulfills the above requirem
e n t ~ . ~ ~ ~ - ~ ~ ~
In addition to single-strand selective probes,
chemicals that react with double-stranded B-.
DNA, such as dimethylsulfate (seeSection V.D.2)
and agents with nuclease activity, can be applied
to DNA structure studies using mainly the footprinting
approach (compakng the reactivities of
DNA itself and its complex with a protein molecule
or some other ligand). Photochemical reagents
represent another class of DNA structure
probes. We confine ourselves here mainly to the
single-strand selective probes that have been
demonstrated to have specific advantages in the
study of local supercoil-stabilized DNA structures.
Other probes are mentioned only briefly.
1. Single-StrandSelective Probes
Table 4 shows the most importantprobes that
react preferentially with single-stranded and some
non-B DNA regions. These probes can be subdivided
into two groups: probes reacting with
sites involved in Watson-Crick hydrogen bonding
and probes reacting with bases at other sites.
The former group includes chloroacetaldehyde
(CAA) (Figure 1c), bromoacetaldehyde (BAA),
glyoxal (Figure Id), and N-cyclohexyl-N'-P(4-
methylmofpholinium)ethylcarbodiimide-p-toluene
(CMC) (Figure le); these probes cannot
react within an intact B-DNA as they require
hydrogen bond breakage (Figures lc to e). The
latter group includes osmium tetroxide complexes,
KMnO,, diethylpyrocarbonate (DEPC),
TABLE 4
Chemical Probes of the DNA Structure
Reacting Preferentially with Single-Stranded
and Non-B-DNA Regions
hydroxylarnine, and methoxylarnine. With the
exception of DEPC, their reactions involve mainly
C5 and C6 of the pyrimidine ring. Nonreactivity
of bases in B-DNA is probably due to steric
reasons.
All of the above-mentioned probesare highly
specific, showing almost no reaction with BDNA,
provided the reaction conditions are properly
chosen. The reaction sites in the DNA chains
can be detected in various ways. Polyclonal antibodies
were elicited in rabbits for the reaction
products of the following probes: bisulfite, 0methylhydroxylarnine
mixture,235CMC,U6 0smium
tetroxide, pyridine ( O ~ , p y ) , ' ~ ~and osmium
tetro~ide,2,2'-bipyridine.'~~.~~'Monoclonals
against the latter DNA adduct were generated in
1990.237
The availability of these antibodies opens up
new possibilities in DNA structure studies, especially
with respect to the detection of open
local structures in situ, and makes possible the
determination of specific DNA adducts at high
sensitivity.
Base
Thus far, osmium tetroxide complexes,
Probe specifity Ref: DEPC, and BAA(CAA) have been applied to the
largest extent to research local DNA structures.
A. Probesreactingwith sitesinvolvedin Watson-Crick Propertiesof these probesare discussed in greater
hydrogen bonding d&ail &I the following paragraphs.
BAA, CAA C,A 268, 286, 287,
288, 290, 291,
308,309 a. Osmium Tetroxide and Its Complexes
Glyoxal G 264, 293,310
CMC T, (3 294
B. Probes reacting in other sites
OsO, (alone)
DEPC
Hydroxylamine
Methoxylamine
NaHSO,
Ozone
Boldnumbers refer to papers containing important
methodological aspects; other represent examples
of the probe application.Further references are given
in text.
Osmium (VIII) tetroxide is a versatile electron
acceptor and an effective reagent for the cis
hydroxylation of alkenes under stoichiometric
conditions (reviewed in Reference 238). In DNA
osmium (VI) esters are formed (Figure 2) via
addition to the 5-6 double bond of the pyrimidine
ring.37.239In the 1970s this reaction was utilized
for the introduction of a heavy metal stain with
the intention of developing an electron microscopic
method of DNA sequencing.240 Ligands
profoundly alter the nature of the osmium tetroxide
reaction, changing the structures, kinetics
of formation, and hydrolytic stabilityof the products.
The reaction of Os,py with monomeric nucleic
acid components have been studied in detail,
chiefly by Behrman et al.239.24'
Os,py reacts most readily with thymine
moieties in single-stranded DNA.38The structure
,
0
- 0
a R
'i'
NC02Et
b
A YH
R
..... CI
4 O
H-C-C,
I H
H
L
c R I
R
C-C,
R
d R
\ ,FIGURE1.
Formation of the adducts between DNA bas& and single-strand
selective probes: (a)osmium tetroxide (alone),(b) diethyl
pyrocarbonate, (c)chloroacetaldehyde,(d) glyoxal, (e)N-cyclohexylN-p-(4-methylmorpholinium)ethylcarbodiimide-ptoluene(CMC),
...,
hydrogen bonding in the Watson-Crick base pair.
of the thymine-Os,py cis ester (Figure 2) was
determined by single-crystal X-ray diffraction
a n a l y ~ i s ; ~ ~ ~ . ~ ~ ~osmium in this ester is approximatelyoctahedral.
Osmiumbindingsites in DNA
chains can be detected in various ways (Figure
3), including detection at single-nucleotide resolution
based either on the labilization of the
sugar-phosphatebackbone (at the site of the thymineosmateester
formation)to piperidinecleavage,38.244-246the
ability of the adduct to terminate
the tran~cription,~~'or the DNA primer extension
in v i t ~ o . ~ ~ ~Adducts of DNA and RNA with various
osmium tetroxide complexes are reducible
at the mercury ele~trode,~~~-=~producing catalyticcurrentsat
negativepotentials;these adducts
can be determined at low concentrations by means
of polarographic (voltammetric) techniques.
In the B-DNA double helix the target C5-C6
double bond of thymine is located in the major
groove where it is not accessible to the bulky
electrophilic osmium probe.231Local changes in
helix geometry may render the C5-C6 double '
bond accessible to the out-of-plane attack of the
Os,py molecule on the base worbitals. Such
changes include base unstacking in the four-way
junction at low ionic strength^,^^.^ single-base
mismatchesand b ~ l g e s , ~ l ' . ~ ~ - " ~local changes in
twist in (A-T), sequences,260helix distortions in
the vicinity of single-strand intermption~,~~~~"
premelting of AT-rich sequences in supercoiled
DNA,26ietc. Os,py has shown an ability to recognize
minute local changes in the DNA molecule
that are not detectable by chemical probes
such as BAA, CAA (both requiring rupture of
II ,
FIGURE 1E
0 0
R R
\
a SO; IY so;
. .
I II m
b
FIGURE 2. (a) Formationof the adductbetween thymine and osmium tetroxide, pyridine;
...,hydrogen bondingin the Watson-Crickbase pair; (b) some ligands which can replace
pyridinein the osmium complex: I, tetramethylethylenediamine(TEMED);11, 2,2'-bipyridine
(bipy); Ill, 1,lo-phenanthrollne(phe); IV, bathophenanthroline (4,7-diphenyl-1,lo-phenanthroline)disulfonic
acid (bpds).
A-
SS
NUCLEASE CLEAVAGE
RII
CLEAVAGE INHlBlTlON
FIGURE3. Schemeof treatmentof supermiledDNA withosmiumtetroxide,pyridine
(Os,py), or 2,2'-bipyridine (Os,bipy) and mapping of the osmium binding sites. To
map the osmium binding sites DNA is cleaved by restrictase (RI) followed (1) by
cleavagewithhot piperidineandsequencingusingtheprinciplesof theMaxam-Gilbert
method; (2) by primer extensionor transcriptionin vitro(terminatedat modifiedbases
in the template) and nucleotide sequencing; (3) by digestion with single-strand selective
nuclease and (nbndenaturing) gel electrophoresis of the resulting DNA fragments;
or (4) if a structural distortion (e.g., in the B-Zjunctiok)'is expected to occur
within the recognitionsite of aparticular restrictase (RII), thesite-specific modification
of thedistortioncanbe manifestedby inhibition of therestrictioncleavage;(5) osmium
binding in plasmids and DNA fragments can be determinedby immunoassay (e.g.,
by DNA gel retardation,immunoblotting,or ELISA). Modificationof large DNA regions
may result in partial or full relaxation of the molecule (accompanied by changes in
the electrophoreticmobility) and formation of a denaturation"bubble"visible by electron
microscopy; the amount of bound osmium can be determinedelectrochemically.
Watson-Crick hydrogen bonds)261(Figure lc), or
DEPC.2'7.z3Within a short time Os,py has proven
to bea usefulprobeof DNA structure in v i t r ~ . ' ~ . ~ ~
At high pyridine concentrations in the Os,py
reagent, low ionic strengths, and long reaction
times, the initial attack of the probe may be followed
by the formation and propagation of the
"denaturationbubble"manifested by changes in .
the DNAelectrophoreticmobility.231.233-234Large
"bubbles" can be visualized in the electron rnic
r o s ~ o p e . ~ ~ . ~ ~ ~Under properly chosen reaction
conditions (e.g, 1 mM osmium tetroxide, 2%
pyridine,in 0.2M NaCI, 25 rnM Tris,5 rnM EDTA,
pH 7.8 15 rnin at 26"C), secondarily induced
changesare negligible;undertheseconditions, sitespecific
modification at the B-Z junctionS245.246.263cruciform
l o ~ p s , ~ ~ . ~ ~ 'protonated
etc. were detected without any
sign of changes induced by the probe. CD measurements
showed no changes in the CD band
in the initial stage of the Os,py reaction with calf
thymus DNA.231Marked changes in CD and a
steep increase in the amount of modified bases
occurred under the: given conditions only after
more than 34 h of the reaction.23'It may thus be
concluded that Os,py is suitable as a probe of
DNA structure, but the reaction conditions must
be carefully chosen if reliable data are to be obtained.
On the other hand, the ability
of pyridine to destabilize the DNA double
helix at higher concentrations may help to detect
some very small changes in DNA stdc-
are.38,230,231.266.267
In Os,py reagent osmium tetroxideis usually
applied at 1 to 2 mM concentrations, while the
concentration of pyridine is higher by two orders
of magnitude. Replacing monodentate pyridine
in Os,py reagent by bidentate 2.2'-bipyridine
(bipy) results in more stable adducts with DNA
and makes it possible to work with osmium tetroxide
and bipy at equimolar concentrati~ns;~~
evensubrnillimolarconcentrationsof Os,bipy are
sufficient for the detection of local open structures
such as the B-Z junctions. In contrast to
Os,py with Os,bipy, no secondary effects were
observed. Os,bipycan be used to probe the DNA
structure in E. coli cells (see Section VTI).
To improve the versatility of the osmium
probes we recently tested different ligands for
their ability to site-specifically modify the B-2
junction and the cruciform loop in supercoiled
plasm id^.^^^.^^^ Ln addition to Os,bipy (Figure 2)
bathophenathrolinedisulfonic acid (bpds) and tetramethylethylenediamine
(TEMED) can site
specifically modify the above-mentioned structures
at millimolar and submillimolar concentrations.
Under the same conditions, a complex of
osmium tetroxidewith 1,lO-phenanthroline(phe)
displayed a lower specificity and also reacted at
other sites of the supercoiled and linear DNA
molecules. Os,phe may thus become useful in
DNA footprintingin vitro and in situ.Differences
in size and other propertiesof Os,bipy, Os,bpds,
Os,TEMED, and Os,phe molecules may come
into play in the selectivity of these probes for
specific DNA structures in vitro and in the probe
penetrationinto the cell and transport to its target
DNA structure, The great diversity in the properties
of osmium (VI) esters offers a number of
possibilities (so far largely unexploited) that are
not available in other chemical probes.
b. Diethylpyrocarbonate (DPC) . ,
DEPC, an enzyme inhibitor and bactericidal
agent, has been applied in nucleic acid research
(reviewed in Reference 271) and protein res
e a r ~ h , ~ ~ ~ . ~ ~ ~due mainly to its ability to react
chemically with some amino acids and nucleic
acid bases. DEPC carbethoxylates N7 of purines
(Figure 1b) in DNA, but under certainconditions
reactions at other sites and with other bases may
o c ~ u r . ~ ~ ~ . ~ ~ ~In RNA, purines in single-stranded
regions are accessible to this reagent, while those
contained in double-stranded regions do not
r e a ~ t . ~ ~ ~ s ~ ~ ~It has been shown that purines in syn
conformation in left-handed 2-DNA react with
DEPC much faster than those in B-DNA.277-278
Enhanced reactivity of 2-DNA toward DEPC
provided a new approach to the studies of the
formation and distribution of 2-DNA segments
within a 2-DNA m o l e c ~ l e . ~ ~ ~ ~ ~ ~ ~DEPCalso reacts
with single-stranded DNA, showing a specific
reaction with bases in the cruciform loop.278.280
As a result of carbethylation and ring opening
(Figure lb) the N-9 glycosidic bond is labilized,
and due to alkali treatment chain scission occurs
at the modified base site. This makes it possible
to detect modified bases (similarly as with Os,py
and 0 s,bipy)at singlenucleotide resolution using
DNA sequencing techniques. Other ways of detecting
the reaction sites in DNA strands have
been little exploited. DEPC has been used for
footprinting DNA-protein complexes;28' however,
this type of experiment should be interpreted
with caution, as DEPC reacts much faster
with proteins thm with' single-stranded nucleic
DNA footprintingis referred to again
in Section V.D.2.
c. Bromo- (BAA) and Chloroaceetaldehyde
(CAA)
BAA and CAA react with adenine and cytosine,283-285forming
fluorescent cyclic ethenoderivatives
(Figure lc). BAA was applied tostudy
local changes in the DNA structure286288using
nuclease S1 to cleave DNA at the BAA-modified
sites. Both BAA and CAA were used to detect
unpaired bases in cruciform loops,287B-2 junctions,268.289and
triplex structures.2w For several
years it was difficult to obtain a pattern at singlenucleotide
resolution by means of the chemical
cleavage at the modified sites. Only recently has
a chemical cleavage procedure been developed
that gives a resolution comparable to that obtained
with DEPC and O ~ , p y . ~ ~ ~It has been generally
accepted that CAA and BAA react only
with bases not included in Watson-Crick base
pairing. Recent results showing CAA reactivity
within certain stretches of Z-DNAz9''inight be
explained by a CAA reaction with a specific
"open" state of the (dC-dA), segments related
to the dynamic equilibrium between B and Z
DNA (see Section VI.B.2).
d. Advantages and Disadvantages
Chemical probes (Table 4) have some advantages
over enzymatic probes:
1. They usually can be applied under a wide
range of conditions, including pH, temperature
and ionic strength values, presence of
nonaqueous solvents, etc.
2. Their moleculesare smaller and can diffuse
to various parts of the DNA structure.
3. Their structures and reaction mechanisms
are known and induction of secondary
changes in DNA structure is less probable
that with large protein molecules.Chemical
probes differ in their specificity toward bases
(Table 4, Figures 1 and 2) and their functional
groups; proper combination of the
chemical probes can thus provide detailed
information about spatial organization of the
"
local DNA structure.
4. They do not induce DNA chain scissions,
making possible the occurrence of simultaneous
reactions at several sites of the supercoiled
DNA molecule.
5. The reaction sites in DNA strands can be
detected in various ways (Figure 3).
6. They are potentially applicable to probe
DNA structure in vivo.
Chemical probes also have some disadvantages:
all chemical probes reacting with DNA
bases are potentiallymutagenic and working with
these reagents may represent a health hazard.
CAAis carcinogenic;284~28sDEPCreactswith arnmonia,
producing the carcinogenic ur15thdne;~'~
OsO, vapors irritates eyes and mucous membranes;
etc. The reagents should be thus handled
with care.
2. Probes.Reacting with DoubleStranded
DNA
There are a number of chemicals capable of
reacting with double-strandedDNA (reviewed in
References 318 to 324) but only those that have
found significantapplication in the study.oflocal
DNA structure research are mentioned here.
a. Dimethylsulfate(DMS)
In B-DNA, DMS methylates N-7 of guanine
and N-3 of adenine, i.e., the shes not involved
in Watson-Crick hydrogen b ~ n d i n g . ~ ' ~ . ~ ~In
single-stranded DNA, N-1 of adenine and to a
lesser extent N-3 of cytosine are also methylated.
In Hoogsteen pairing N-7 is involved in
hydrogen bonding; it does not react with DMS.
This makes DMS suitable for probing structures
involving Hoogsteen pairing, such as protonated
triplex DMS alsoyields information
about ligandlguanine contacts in the major
groove of B-DNA, where N-7 of guanine is lo-
cated.329-331
N-Ethyl-N-nitrosourea (ENU) reacts with
DNA backbone phosphates, forming triesters. The
reaction sites can be recognized by the triester
alkaline hydrolysis.332Ethylation of phosphates
has been used for interference studies of ligand
interactions with DNA phosphates.332J33The
technique does not detect direct contacts in the
DNA complex, but rather that phosphates that
are located too close to the ligand to accommo-
date
the ethyl group. The conditions of the reaction
are far from physiological, thus significantlylimiting
the useof ENU in DNA studies.332
Modification of DNA with N-methyl-N-nitrosourea
(MNU) under physiologicalconditions
-showed that the initial attack of MNU is strongly
dependent on DNA conformation, suggesting that
in addition to phosphates, bases might be involved
in the rea~tion.~"
c. Probes with "Nuclease" Activities
Some chemicals are capable of inducing DNA
chain scission, thus simulating to a certain extent
the behaviorof natural nucleases. ~ucleasessuch
as DNaseI, DNaseII, and micrococcal nuclease
were applied in DNA f ~ o t p r i n t i n g , ~ ~ ~ . ~ ~ 'i.e., the
technique used in the analysis of specific protein
and drug binding to DNA. In footprinting, DNA
fragments are cleaved with a nuclease and the
products are analyzed using the Maxam and Gilbert
technique,335in which the sites protected
with specifically bound protein or drug molecules
are observed as gaps in the DNA cleavage pattern.
In addition to natural nucleases, chemicals have
been applied recently to cleave DNA, including
1,lO-phenanthroline-copperion,322methidiumpr~pyl-EDTA,~~and
Iron @).EDTA.320With the
two latter chemicals DNA reacts to generate the
hydroxyl radical (.OH), which induces the chain
cleavage.In contrast to methidiumpropyl-EDTA,
Iron (II).EDTA does not bind to DNA, which is
cut by freely diffusible .OH. With Iron(II).EDTA
footprint at very high resolution can be obtained.
On the other hand, specific binding of phenanthroline-copperto
B-DNA can be,.utiiizedin DNA
structure studies.322This chemical cleaves A-DNA
'
slower than B-DNA, while Z-DNA and singlestranded
DNA are n; cleaved under the same
conditions. More details can be obtained in References
332 and 336 to 341.
d. Photochemical Probes
Psoralens intercalate within the double helical
DNA and on irradiation they are covalently
added to the thymine 5,6 double bond.342One
psoralen molecule can photoreact with two thymines
at nucleotide sequences containing adjacent
thymines in opposite strands, creating a DNA
interstrand crosslink. The crosslinking sites can
be analyzed by electron microscopy. To improve.
the limited resolution of this technique, enzymebased
procedures were recently developed that
make it possible to analyze sites of psoralen adducts
at single-nucleotide res~lution."~-~
Barton et al.324."7-350developed a series of
transition metal complexes that bind specifically
to local DNA structures; their photoactivation results
in a site-specific DNA strand cleavage. For
example, the tris (phenanthroline) metal complexes
are favored for intercalation into righthanded
helices, while the bulkier isomer of tris
(4,7-diphenylphenanthroline)ruthenium binds to
left-handed 2-DNA or to any other conformation
that is unwound sufficiently to accommodate the
bulky complex. Rhodium(1II) complex binds to
the cruciform and upon photoactivation it cleaves
DNA at and near the ~ruciform."~A probe for
A-DNA was developed by matching the shape
of the ruthenium(II) complex to that of the shallow
minor groove of the A-DNA f o r ~ n . ~ ~ . ~ ' ~
e. Complementary Addressed Modification
and Cleavage of DNA
The ability of DNA strands to renature and
hybridize was recognized more than 30 years
ago.351-353Utilizationof this principle has become
one of the main driving forces in the development
of specific molecular biological techniques, of
which one is the technique of complementary
addressed modification and cleavage of DNA. Here,
a relatively short oligonucleotide equipped with a
chemi~allyrpctive group, intercalatingagent, andl
or nucleic acid cleaving agent binds to its complementary
target sequence in single-stranded DNA
or RNA (reviewed in References 354 to 359). In
double-stranded DNA, sequence-specific modification
with alkylating oligonucleotides was observed
in a negatively supercoiled plasmid but not
in its relaxed form.360The double helix in relaxed
DNA molecules can, however, be a target for pyrimidine
(or purine) oligonucleotidesthat recognize
homopurine~homopyrimidine(homopu-py) sequences
via the major groove and form triplex
~ t r ~ c t u r e ~ . ~ ~ ~ - ~ ~ ~ . ~ ~ ~ - ~ ~The oligonucleotide is
bound in a parallelorientation to the purinestrand
of the DNA double helix.
An oligonucleotide of about 15 to 19 nucleotides
should be sufficiently long to recognize
a single specific sequence in the human genome.356This
technique may be of great use for
mapping genomes over large distances to study
local DNA structures, namely, the DNA triplexes
both in vitro and in vivo, etc.
While in experiments in vitro the usual deoxyoligonucleotides
can be applied, for ex-
perimentsin vivo several modificationshave been
introduced into oligonucleotides to prevent their
degradation by nucleases. The phosphate group
was replaced by methylphosphonate and phosph~thionate,~~~and
the a-nucleotide anomers were
substituted for the natural p-anorner~.~~~The covalently
linked intercalating agents include acridine,
phenanthroline, proflavin, and furocoumarin
derivatives.357.366-368 Proflavin and
furocoumarins are photoactive and can be attached
to the oligonucleotides by irradiation.3s6
As a nucleic acid-cleaving agent, metal complexes
such as Iron-EDTA, Iron-porphyrin, and
1,lO-phenanthroline-copperhave been used.356
Oligonucleotide binding can be irreversible
when the covalent bond is formed. An irreversible
and well-detectablechange is a site-specific
DNA chain scission induced by a DNA-cleaving
agent such as Iron(II)*EDTAattached to the olig
~ n u c l e o t i d e . ~ ~ ~ . ~ ~ ~ ~ ~ ~Recently, an irradiationinducedcleavagewas
reportedwith an I1-residue
homopyrimidine oligonucleotide covalently
linked to ellipticine derivative that introduced
cleavage of the two strands of the target homopu-py
sequence.369Development of the sequence-specific
artificial photonucleases of this
type represents a new approach in which the
cleavage reaction can be very well controlled by
light.
Oligonucleotideconjugates that bind specifically
to mRNA inhibit its translation in cell-free
extract^.^^^,^^^ The use of antisense oligonucleotide
to selectively control gene expression is a
very promising approach of potentially great significance
that may provide the basis for the development
of highly specific therapeutic agents.
At present, however, some basic questions remain
unanswered.370
Site-specific DNA chain scissions have also
been accomplished by incorporation of DNA
cleaving agents into sequence-specific DNA
binding peptides and proteins, e.g., the E. coli
catabolitegene activator, a helix-turn-helixmotif
sequence-specific binding protein.371
f. Ultraviolet Irradiation
The formation of UV light-induced cyclobutane
pyrimidine dimers in DNA requires a
proper alignment of the 5.6 double bond in the
two reacting pyrimidines.This requirementis not
fulfilled in B-DNA; thus, the dimer formation
can serve as a DNA structural probe.320.372J73
Formation of the interstrand crosslinking dimers
was used more than 20 years ago to study DNA
premelting,10.375.376but in the years hence, this
approach was little exploited. Only recently was
a renewed interest in this technique shown. It has
been demonstrated that peaks of pyrimidine dirner
formation occur in nucleosomecore particles
with an average periodicity of 10.3 base^,^".^^^
and that the rate of thymine dimer formation is
affected by the direction and degree of DNA
bending.379Homopyrimidine inserts [with (dTdC),
or (dC), sequences] in plasmid DNA are
good targetsfor UV-induced[6-41-pyrimidine dimer
formation,380as demonstrated by photoprints
of fragments produced by DNA piperidinecleavage.
The dimerization in these sequences was
almost completely abolished when homopyrimidine
oligonucleotides with (T-C), or (C), sequences
were added to form the triplex structure.
This technique may become very useful in the
research of DNA triplex structures, both in vitro
and in vivo.
UV irradiation has also been applied as a
photocrosslinkingand footprinting agent in studies
of protein-DNA interaction^.^'^.^^^"^'^^^^ Pulsed
laser (nanoseconďpulses) and flash irradiation
(microsecond pulses) have been increasingly applied
for time resolution ~ t u d i e s . ~ ~ ~ ' ~ ~ 'It appears
that thymines photoreact with primary mines,
including lysine residues, in proteins, but the
photochemistry of protein-DNA crosslinking is
not yet fully understood.
3. Conclusions .. .
In contrast to the early 1980s, when the arsenal
of methods suitable for the analysis of local
supercoil stabilized structures was rather limited,
a large number of methods are currently available.
Among them chemical probes and gel electiophoresis
are perhaps the most important. In
studies of oligonucleotides modeling a specific
DNA structure, physical techniques and particularly
NMR386can yield detailed structural information.
It can be expected that scanning Nn-
neling microscopy, whose usefulness in the
DNA structure studies was recently de'fnon~
t r a t e d , ~ ~ ' - ~ ~ ~will soon become a very useful tool
in the research of local DNA structures and their
interactions. Attempts to improve our knowledge.
of DNA structure inside the cell will require the
development of new techniques. The first results
in this respect have been obtained with chemical
probesand specific moleculargenetic approaches
(see Section VLI). It is anticipated that further,
rapid development of these techniques will continue
in the near future....
VI. LOCAL CHANGES IN DNA
SECONDARY STRUCTURE STABILIZED
BY SUPERCOILING
According to Equation 1, supercoiled DNA
with ALk # 0 must either twist or writhe, or
both, away from its relaxed form. Changes in
=may involve torsional deformations of existing
s e c o u struc- and/or transitio~sof BDNA
segments t~ supercoil-stabilized specific
local structures depending on nucleotide sequence,.These
transitionsinclude B-Z transitions
at alternating purine-pyrimidine sequences, hiplex
formation at homopwpy tracts, cruciform
extrusions at inverted repeat sequences, etc.;
changes in JQinvolve bending and the formation
of superhelicalturns. It should be noted that local
changes resulting from conformational transitions
must be compensated for elsewhere in the
molecule (supercoiled domain) to preserve the
linking number.
Thus far the best understood structural change
in a negatively supercoiled DNA molecule may
be the cruciform e ~ t r u s i o n , ~ ' ~ ~ ~ ~ . ~ ~ 'discussed in
detail in Section V1.A; here, it is mentioned as
an example of formation of a local structure assisted
by freeenergy of the negativesupercoiling.
Inverted repeat sequences can, in principle,
undergo a transition to a structure in which the
regions with intrastmnd complementaritycan.pair,
forming the cruciformstructure. Such a structure
is energetically unfavorable compared with perfectly
base-paired continuous DNA duplex. Thi
energetic disadvantage of cruciform formation,
however, can be overcome if the inverted repeat
is placed in the negatively supercoiled molecule,
in which the free energy of superhelix formation
can provide the driving energy required to facilitate
the cruciform extrusion. An inverted repeat
of, for example, 21 bp, will in its unperturbed
state contribute to the total Tw by 21110.5 or 2.
The inverted repeat contribution will decrease to
zero after the cruciform extrusion provided the
backbone conformation adopted within the hairpin
itself is relevant to the topological properties
of the whole molecule.'22A similar decrease in
Tw can be expected if the same sequence forms
a melted region, while formation of 2-DNA would
result in a DNA unwinding almost twofold
greater.
In natural supercoiled DNAs usually more
than one nucleotide sequence in the molecule is
able to undergo transition to a specific local supercoil-stabilized
structure. Once the first transition
occurs, the probabilityof further transition
is changed-due to the partial relaxation of the
molecule induced by the first transition. Moreover,
a single nucleotide sequence may itself be
able to form several alternative structures, depending
on various factors, including environmental
condition^,^^.^^^ superhelix density,z3etc.
Thus, complex multistate equilibria may arise in
supercoiledDNA molecules;such equilibria may
thke part in DNA-controlled regulatory processes
in vivo.
Presence of the local structuresin supercoiled
DNAs complicates their geometric and topological
analysis. Equation 1 was formulated in
terms of winding of either DNA strand about the
duplex axis that is often difficult to define. To
overcome this difficulty, new models were proposed
(see Section IV) capable of evaluating
quantitatively complex phenomena connected
with the presence of local structures such as cruciform~in
supercoiled DNA molecules. In this
section conformations and properties of cruciforms,
left-handedsegments, triplexes,and other
local structures in vitro are discussed. The question
of the existence of these structures in vivo
and their biological relevance is a subject of Section
VII.
A. Cruciforms
Cruciform structure was predicted 35 years
ago,,393s39but the importance of the negative sup&coiling
for its formation was suggest=d and
theoretically derived much The first
experimentalevidence of the cruciform structure
was obtained about 10 years ago with the observation
of very large cruciforms in ,the
electronmicroscope2sand by nucleaseS1 probing
of supercoiled plasmids and phage
The possibility that the structure
might be artificially induced by acid pH and
Zn2
+ ions of the enzyme probe buffer and/or by
interaction of the enzyme molecule with
DNA was soon excluded by chemical probing234
and 2-D gel electrophoresi~'~~.~~~at neutral pH
in the absence of Zn2
+ ions. Cruciform is now
well established and various aspects of its stmcture
were reviewed in a number of papers.31.35.122.123.138,390.391.400
The formation of a
cruciform requires a nucleotide sequence with a
local twofold symmetry represented usually by a
sufficiently long inverted repeat. Two structural
features are characteristic for the cruciform, i.e.,
the formally unpaired loop and the four-way
junction.
1. Loop
Optimal size of the formally single-stranded
loop lies between four and six bases.79.401.402The
loop adopts some structure involving probably
base sta~king~~~.'"'~and even non-Watson-Crick
base pairing.402Hairpinloops may affect the stem
conformationclose to the loop with the stem perturbation
stronger for smaller loops.404
2. Four-Way Junction
The four-way junction3
' is normally fully base
paired,79.279.m.m5asymmetric and X-shaped.
It introduces a bend into the molecule whose
nature differs from the intrinsic sequence-directed
c u r ~ a t u r e . ~ ~ . ~Ln the absence of ions thymines
at the junctionare reactive to Os,py, probably
due to base unsta~king,~and the helical
armsof the junctionare fullyextended in a square,
configuration. The four-way junction is formally,
equivalent to the Holliday junction - a fundamental
structure in the genetic recombination. A
series of papers was published in 1989 and 1990
by Lilley et a1.*ď4 demonstrating for the first
time the structural details of the Holliday junction.
Their experimental approach was based on
hybridization of four oligonucleotides(each of 80
nucleotides) to assemble a four-way helical junction.
This complex and its variants (containing
small changes in nucleotide sequence or substitution
of electrically neutral methylphosphonate
at the central phosphodiester linkages) were investigated
using various techniques, including gel
electrophoresis,chemical probes, nucleasecleavand
the resulting model was confirmed
and refinedon the groundsof fluorescenceenergy
transfeFL4and model building studies.411The
model is called the stacked X-structure in which
the four arms are associated in pairs, stacked in
a coaxial manntr' ťt9generate quasi-continuous
helices. The nucleotide sequence at the junction
center determines the choice of the stacking partners.
Binding of metal ions is important for the
stability of the X-structure. Resolvase (from T4
phage) binds at one side of the X-structurecleaving
the exchanging strands two or three nucleotides
from the junction. More detailscan be found
in Lilley and co-workers' recent
Information about the details of the cmciform
loop, st&, arid four-way junction was obtained
not only by investigation of supercoiled
DNAs by techniques specific for local supercoilstabilized
structures (this section), but also by
studying oligonucleotide constructs modeling
cruciform structure or its specific
parts401.402~404.405.407-416 by techn
i q u e ~ , ~ ' . ~ ~ . ~ . ~ ~ . ~ ' ~namely, NMR. Using this
approach a crystal structure of a hexadecanucleotide
CGCGCGTTTTCGCGCG was solved at
2.1 A resolution showing a hairpin configuration
with Z-DNA hexamer stem.416Ln contrast to this
type of research studies of the cruciform extrusion
had to be done with supercoiled DNA molecules
and the applied techniques were thus limited
mainly to those described in Section V.
3. Cruciform Extrusion
Extrusion of the cruciform requires significant
rearrangement of the DNA structure, including
complete reorganizationof the base pair-
ing. Compared with the regular DNA duplex,
cruciform is thermodynamically uristable due to
the energetic cost of base unpairing in the loop
and formation of the four-way junction. Free energy
of the cruciform formation was determined
to be 15 to 20 kcal/mo1.'38.172-399.417
The first studies demonstrated a significant
kinetic bamer and showed that the cruciform extrusion
could be a rather slow p r o ~ e s s . ~ ' ~ . ~ ' ~Further
work withother DNAs revealed,- however,
that extrusion could proceed under some conditions
more easily. Detailed studies showed two
types of cruciformextrusion kinetics. Cruciforms
requiring the presence of salt for their extrusion
(showing maximum extrusion rates at about 50
mM NaCl) are termed wg(related to salt dependency).
S-type cruciforms are more common
than C-type crucifom, so far represented only
by ColE1as the only natural memberof the Ctype
(C for ColEl) class. C-type cruciforms extrude
maximally in the absence of salt, showing a
marked dependence on temperature, which is in
contrast to the ratherlow temperaturedependence
of theextrusion of theS-type cruciforms.Striking
differences in the kinetic properties of S- and Ctype
cruciforms suggested the possibility of two
different mechanistic pathways for the extrusion
process (Figure 4). The S-type extrusion is idtiated
by a helix opening at the center of the
inserted repeat, followed by the formation of a
small protocruciform and branch migration resulting
in elongation of the cruciform stems up
to the complete extrusion. Such extrusion was
explained theoretically by Vologodskiiand FrankKarnenet~kii~~~and
documented by a wealth of
experimental evidence (Reference 391 and references
therein). Recent suggest that the
initial opening may correspond to about 10 bp
and that the nucleotide sequence at the center of
the inverted repeat significantly influences the
rate of cruciform High A+T
content of this region or destabilization of the
helix by methylation of adenine at N6 enhanced
the extrusion rate, while an increase:in the G+C content
or stabilization of the helix by cytosine
methylation resulted in a decrease of the extrusion
rate. These results, as well as cation binding
analysis, support the concept of the S-type cruciform
extrusion.426
The C-type extrusion (Figure 4) is supposed
*.em
to start with coordinate unpairing of a much larger
duplex segment followed by intrastrand reassociation
to form the fully extruded cruciforq. This
pathway would be expected to be facilitated by
low ionic strength and characterized by a large
activation energy, i.e., the characteristics experimentally
observed, but what are the reasons for
the two different extrusion mechanisms? Comparison
of the nucleotide sequence of ColEl
cruciform with that of a typical S-type cruciform
does not show differences that may help
answer this question. The explanation of the
C-type extrusion offered recently by Lilley
et al.31-26'.424.42u29is based on the presence of
very (A+T)-rich sequences Qanking the ColEl
inverted repeat. Replacing ColE1 cruciform with
another inverted repeat (originally S-type) resulted
in a typical C-type extrusion429Similarly
replacing an S-type cruciform in its natural environment
with a ColEl inverted repeat exhibited
S-type extrusions. This clearly demonstrated that
for the C-type extrusion, the sequences flanking
the ColEl repeat are responsible.
It was further shown that
1. The transition energy of the cruciform extrusions
decreases with the length of the Ctype
indbcing sequence43'
2. The polarity of the sequence can be unimportant
and its length can be reduced to
12427.430
3. If the inverted repeat is flanked both by Sand
C-type sequences, the salt concentration
dominates the type of kinetics (i.e., Ctype
in absence of NaC1)
4. The effect may be manifested at a significant
distance (around 100 bp)
5. Insertion of a (G+C) segment between the
C-type inducing sequence and the inverted
repeat may block the C-type extrusion
How can the (A +T)-rich C-type inducin4
sequence,perform a remote control of the cru:
ciform extrusion? It was proposed that these sequences
are responsible for a coordinate destabilization
of a large domain in the supercoiled
DNA, thus increasing the probability of largescale
base opening in the inverted repeat.3t.261.390-427In
agreement with this proposal,
duplex stabilization in (A-T)-richsequenceswith -
A
cooperat lve
/ -opening
large bubble
/ C-type extrusion \
inverted repeat
fully-extruded
\ S-typeextrusion L 2 r u c " o r mcentral
cpenlng
migraticn
intra- s trond
polring
-- n
central bubble proto-cruciform
FIGURE 4. Mechanism of cruciform extrusion. The inverted repeat, represented
by the thicker line, is shown in the unextruded form on the left. C-type
cruciforms(top) initiate theextrusionprocesswith acoordinateopeningof many
base pairs to form a large bubble. An intrastrandreassociation then forms the
mature cruciform structure. The extrusion of S-type cruciforms (bottom), is
initiated by a smaller opening event. lntrastrand pairing generates a smaller
protocruciform, which may undergo branch migration. Base pairing is transferred
from unextruded sequence to the growing cruciform stem in a multistep
process to form the fully extruded structure. The principal differencesbetween
the two mechanisms lie in the initial opening and the degree of tertiary folding
in the transition state. (From Lilley, D. M. J., Chem. Soc. Rev., 18, 53, 1989.
With permission.)
--
distamycin and NaCl resulted in S-type extrusion.
Conversely, helix destabilizing agents induced
in DNA with S-type extrusion a quasi-Ctype
extrusion kinetics. Thus, changing the helical
stability results in interconversion between
S-and C-type mechanisms.
Bases in the C-type inducing sequence in
supercoiled DNA are hyperreactive toward Os,py,
BAA, glyoxal, NaHS03,261and DEPC.431This
hyperreactivityrequires a threshold level of negative
supercoiling and changes in the chemical
reactivity induced by various agents correlate with
those in cruciform extrusion.261BAA hyperreactivity
was detected at temperatures above 26°C
(with a maximum around 40°C), while that of
Os,py was observed above 0°C reaching a max:
imum of 20°C. The BAA hyperreactivity was
explained by a cooperative but transient helix
opening. Os,py hyperreactivity at much lower
temperaturesmight be due at least in part to helix
destabilization by 3% pyridine at low ionic
s t ~ e n g t h . ~ ~ ~
, P
' ~ypersens6ikyToward BAA, DEPC, and
KMnO, was observed in a "random sequence"
region centered 10 to 30 bases away from the
junction of the (A-T),, cruciform;432this hypersensitivity
was pH-dependent. (A-T), sequences
adopt the cruciform structure at low energies of
formation without a detectable kinetic barrier.173.288.433
The unusual extrusion kinetics of
the (A-T), cruciform may be due to specificproperties
of (A-T), duplexes that are very easily deformable
and denaturable and whose structure - .._
differs from the regular B-DNA.260.4"It has been
suggested that these (A-T), sequences may represent
simultaneously the inverted repeat and an
effective inducing sequence, being thus a special
subclass of the C-type extr~sion.~'The hypersensitivity
of the region in the vicinity of the (AT),,
cruciform was related to the cruciform extrusion,
but its relation to the C-type extrusion
remained unclear.
The C-type cruciform extrusion is a very interesting
phenomenon and some of its aspects
may have more general consequences. h was
shown quite recently that the AT-rich flanking
sequences may influence the structure of lefthanded
DNA (see Section VI.B.2).43s At the
present time it would be desirable to have more
detailed information on the way in which the
inducing sequence materializes the remote control
of the cruciform extrusion. The suggested3
'
telestability e f f e ~ t 4 ~ ~ ~would at least requiresome
considerationsconcerning DNA supercoilingand
the explthation based on the soliton-like states
in DNA seems to be even less a ~ c e p t a b l e . ~ ~ ~ . ~ ~ ~
There is no doubt that the ability of some DNA
segments to induce local changes in DNA conformation
far away from the segment may be of
great biological importance. Their better understanding,
however, requires further work.
4. Multiple Structures in Inverted
Repeats
Some inverted repeats may be composed of
alternating purine-pyrimidinesequences and can
(in principle)form either a cruciform or a 2-DNA
structure. It was shown that the
(CATG),,*(CATG),,. insert selectively forms a
cruciform structure when integrated in a negatively
supercoiled p l a ~ m i d . ~ ~ ~Similarly,
(TCGA),-(TCGA), and (TG),.(CA), inserts preferentiallyadopted
cruciformat low ionic strength
rather than Z-DNA?' However, the latter two
sequences were induced to form left-handed DNA
(after removal of local structures by ethidium
bromide) when the negative supercoiling necessary
for the transition was generated at higher
ionic strength (e.g., 0.2 M NaCI).
Although 2-DNA formation in (C-G) inserts
requires more free energy than cruciform formation
these sequences preferred the left-handed
form.226(GTAC), flanked with two (G-C), sequences
adopted left-handed DNA.w1(A-T) sequences
showed a strong preference for 'crukiform
f o r r n a t i ~ n , ~ ~ . ~ ~ ~but under specific conditions
(in the presence of Ni2
+ ions) left-handed structure
was observed in a supercoiled plasmid (see
Section V1.B. These results suggest that
equilibria between B, 2, and cruciformstructures
exist in alternating puepyrsequences depending
on thesuperhelicity andenvironmentalconditions.
The possibility that a given sequence may
undergo more than one structural transition may
be of great biological significance as these transitions
provide specific recognition sites (loops
and four-way junctions of the cruciform and B-
2 junctions and altered helical sense of 2-DNA)
for DNA binding proteins and cause different
levels of DNA relaxation, which may transmit
long-range structural effects influencing regulatory
processes.260p392.440
B. Left-Handed Z-DNA
About 8 years after publication of the CD
spectra of the left-handed DNA in poly(dGdC)-poly(dG-dC)by
Pohl and J o ~ i n , ' ~ . ' ~2-DNA
structure was solved by single-crystal X-ray diffraction
studies (see Section II).'9-21This discovery
stimulated enormous scientific effort, resulting
in hundredsof papers that have been surveyed
in numerous reviews.3636,138.191.w2447It would be
useless to try to present a more or less comprehensive
review of the broad field of 2-DNA in
this article; instead I shall touch on some problems
studied in recent years that have not been
'fully covered in thti preceding reviews. These
include the question of formation of left-handed
DNA within (dA-dT), sequences and the B-Z
junctions.
1. Left-Handed DNA in (dA-dT),
Sequences
Regular alternations involving anti and syn
nucleoside conformations are one of the main
characteristics of Z-DNA (see Section II). Such
alternationsare formed most easily in alternating
pu-pyrnucleotide sequences (with purines in syn
conformation). (dC-dG), sequences easily adopt
the 2-DNA structure, as has been shown both by
solution and crystal studiesw2 The surprising reluctance
of the corresponding (dG-dC), oligonucleotides
to adopt 2-DNA structure has not yet
been fully explained. The ability of (dA-dC);(dTdG),
sequences to form left-handed helices in
solution both in lineaS48-J5'and supercoiled
D N A s ~ ~ - ~ ~ ~ . ~ ~ . ~ ~ ~is well known, but no crystal
structure has been reported up to now.
Consecutive AT pairs can be incorporated
into the Z helix (reviewed in References 442 and
445) and a maximum of six AT pairs can be
tolerated, including the (T-A), sequence. In longer
(dA-dT) sequences no left-handed structure was
observed under conditions where (dC-dG), anď
(dT-dG);(dA-dC), sequences adopted this struc--
ture. Inversion of the CD spectra of poly(dAdT)-poly(dA-dT)in
concentrated CsF solutions
was not due to the B-Z transition but to the socalled
B-X t r a n s i ~ i o n . ~ ~ ~ . " ~ ~
The frrst evidence of the B-Z transition was
obtained in 1986 in Tallandier's laboratory by
investigating poly(dA-dT).poly(dA-dT) in films
by IR spectroscopy in the presence of different
counter ions and a wide variety of water cont
e n t ~ . ~ ~ ~IR spectra observed in the presence of
Ni2
+ at high polynucleotide concentration and
low water activity were assigned to a Z-type
structure of poly(dA-dT).poly(dA-dT). CD of
poly(dA-dT)*poly(dA-dT)solution at high NaCl
concentration(5 M)in the presence of NiCI, (90
rnM) displayed almost total spectrum inversion
with no negativeband around 250 nm, suggesting
Z conformation of the polyn~cleotide.~~~The
nickel-induced Z-form of poly(dA-dT).poly(dAdT)
obtained further support from UV
absorption456and especially from Raman457and
resonance Raman measurements.458It was proposed
that poly(dA-dT).poly(dA-dT) forms ZDNA
due tointeractionof nickel ion with adenine
N7.457*458This interaction is possible thanks to
the screening of negatively charged phosphates
by high sodium concentration that results in stabilization
of adenosine in syn conformation and
reorganization of the water distribution along the
molecule that stimulates the B-Z transition. At
NiCI, concentrations close to 0.1 M pH cannot
be neutral, but it must be weakly acidic. The role
of acid pH in the formation of Z-DNA in (dAdT)
sequences, however, was not considered.
Recently it was shown392that (dA-dT),, insert
can adopt left-handed structure in a supercoiled
plasmid under conditions (2M NaC1,0.2 M
NiCI,, or 1 M NiCI, alone) not sufficient to
induce Z-DNA in the linear plasmid or poly('dAdT)-poly(dA-dT).
(dA-dT), block placed in the
center of the 32 bp~c5mplementaryaliernating
purine-pyrimidine insert adopted a left-handed
structure459at substantially lower NiCI, concentrations
(10 mM NiCI,, 0.2 M NaCI) as detected
by DEPC probing. These results suggest in agreement
with the recent predictions that all simple
alternating purine-pyrimidine sequences such as
G-C, A-C, and A-T may adopt left-handed conformation
under some environmental conditions
in linear D N A S ; " ~ . ~ ~in supercoiled DNAs these
conditions must be combined with a proper superhelix
density.
2. 6-Z Junctions
Segments of left-handed Z-DNA can be contiguous
with B-DNA both in vitro and in vivo
(see Section VLI). The boundary between these
two structures has been specified as B-Z junction
(reviewed in References 34,36,442,444 to 446,
and 460). Structural models of the B-Z junction
suggest that at least one base pair at the junction,
must be different from the B or Z conformat
i o n ~ . ~ ~ ' - ~ ~Experimental data obtained mainly
with supercoiled plasmids show that the B-Z
junction may be a rather polymorphic structure,
Cleavage of the B-Z junctions in supercoiled
plasmids with single-strand selective nucleases
(such as nuclease S1) as well as site-specific
modification with single-strand selective chemic$
probes (see Sections V.C and V.D. l), including
0 ~ , ~ ~ 2 4 5 . 2 6 8 . 3 9 3 . 4 6 6 - 4 6 8 . 4 7 0and hydroxylamine,245.468indicated
open distorted structures,
but the data could not be unambiguously
interpreted in terms of a presence of unpaired
bases (not required in the structural models),
as the above nucleases and chemicals can
interact even with distorted base-paired regions.
To detect unpaired bases in the B-Z junction
chemicals reacting specifically with sites involved
in Watson-Crick hydrogen bonding were
applied.289.291,308.466.471
In 1987 it was shown268.Xl.293.308.J70n471 that
such chemicals, including BAA, CAA, and
glyoxal, site-specifically modify the boundary
between B- and Z-DNA formed in supercoiled
plasmidseither by (G-C),268-293.308.47'or (T-G)n471
sequences. This implies that the B-Z junctions in
supercoiled DNA contain unpaired bases. On the
other hand, further conclusions regarding, for example,
the exact number of unpaired bases, must
be considered with caution because the nuclease
S1 applied for the detection of BAA or CAAmodified
sites may not recognize"solitary nucleotides
modified by the probe.268-293TO obtain
more accuratedata, the recently developed method
of chemical cleavage at the sites of BAA modification
should be applied.290
No crystallographic data have thus far been
published describing structure of the B-Z junction.
Recently, however, other physical
technique^^^^-"'^ were applied to study the B-Z
junction in a linear DNA fragment472and olig
o n u c l e o t i d e ~ . ~ ~ ~ -~ ~ ~It was shown that the hydrddynamic
dimensions of the 153 bp fragment
are not affected by the transition of two (C-G),
segments to Z-DNA, induced by 15 mM
[CO(NH,),]~+.~~~This may imply that the B-Z
junctions were, under the given conditions, neither
strongly bent nor particularly flexible. On
the other hand, the results of IRspectroscopy of
nucleotide films suggest that flexibility in the
junction region (resulting from the presence of
one no-base residue)stimulates the B-Z transition
in six nucleotide residues of a double-stranded
tride~amer.~'~
Recently, a 16 bp oligonucleotide
"C G " C G " C G " C G A C T G A
G " C G " C G " C G " C T G A C T
1 2 3 4 5 6 7 8 9 1 0 1 1 12'13
was synthesized with 5-methylcytosine ("C) incorporated
in the (C-G), segment, which adopted
a Z-conformation at high (>3 M)NaCl concentration~.~'~Studies
of this oligonucleotideby UV
absorption and CD and NMR spectroscopy revealed
three base pairs involved in the B-Z junction
with only one of them being dramatically
d i s t ~ r t e d . ~ ~ ~ . ~ ~ ~The proton resonances for base
pairs 7,8, and possibly9 became observableonly
at temperatures higher than approximately 30°C,
suggesting that at lower temperatures the base
pairs are accessible to exchange with solvent. At
30 to 50°C, all internal hydrogen bonds in base
pairs 2 to 14 were intact. It has been shown that
methylationof a plasmid with HhaI methylase in '
vitro decreases unwinding in the B-Z junction by
half as compared to the unmethylatedp l a ~ m i d . ~ ~ ~
The question thus arises to what extent are the
structural features of the B-Z junction observed
by NMR in the tridecamer influenced by cytosine
methylation.474
The results of chemical probing of the
B-Z junctions in supercoiled plasmidS246.263.268.289.293.30884W471
suggest that the B-Z,
jun~ti~nsadjacent to (C-G), segments are narrow
in good agreement with the results of the oligonucleotide
proton NMR.474B-Z junctions adjacent
to (G-T), segmentsin supercoiledplasmids
may contain more bases hypersensitive to the
chemical probe and might be structurally different.268.467470It
was shown that in (dC-dA),, sequence
around a -0.06 only about half of the
62 bp tract exists in Z-form anywhere along the
C M G segment and is probably constantly in
m ~ t i o n . ~ ~ ~ . ~ ~ ~ ~ ~Under these conditions the hypersensitivity
toward chemical probes associated
with B-Z junctions can be distributed over many
bases. At more negative a the junction hypersensitivity
begins to appear at one end of the Zforming
sequence, suggesting that the structure
of a B-Z junction can be assumed at less energy
costs at one boundary than at the other.
Do these data suggest that the left-handed
structure of (T-G);(A-C), segments is different
from Z-DNA observed in (C-G), sequences?The
C T G
G A C ,, .\:
14 15 16 L 5
available experimental data do not exclude such
a possibility;some of them [e.g., the free energy
of formation of Z-helices in (T-G);(C-A), sequences
is higher than for (C-G), stretchesof the
same length] might even support it.J45Quite recent
results of Rajagopalan et al.435makes the
possibilityof the existenceof various left-handed
structures even more probable. These authors
showed that flanking of Z-forming (C-G) sequences
with AT-rich ColEl sequences(see Section
VI.A.2) resulted in site-specific Os,bipyand
BAA modifications within the left-handed (C-G)
tracts of 36 and 40 bp in length. To explain this
unexpected result it was suggested that the (CG)
tract adopted two conformations that created
a new junction.
Regarding the B-Z junctions, at the present
time we may only conclude that their structure'
is polymorphic depending on nucleotide sequence,
superhelical density, and environmental conditions.
Bases in the junction show not only
hypersensitivity to single-strand selective chemicals
(Table 4). but also represent sites of enhanced
intercalation of the psoralen probe.447
Structureand propertiesof the B-Z junctionsmay
profoundly influence the B-Z transition, and with
their open structures differing markedly from
those of the adjacent helices, these junctions
themselves may play a significant biological role
(representing, for instance, a binding site for single-strand
binding protein^).^'^.^'^ Further work
will be,needed, including X-ray crystallography,
NMR, and other physical techniques, as well as
chemical probing to understand better the structural
features of the B-Z junctions.
C. Triplexes
In only a few years triplexes have become
one of the most studied aspects of the DNA polymorphic
structure. ~ o l e c u l a gtriplexes at
weakly acidic and neutral pH values require supercoiling,
while intemolecular triplexes from
duplex DNAs and oligonucleotidescan be formed
under the same conditions in relaxed DNA molecules.
As this article deals with supercoil-stabilized
structures, I shall focus on intramolecular
triplexes in an attempt to present a more detailed
review. Intermolecular complexes are only briefly
summarized.
1. Early Studies of Multistranded
Structures in Synthetic Polynucleotides
More than 30 years ago poly(A) and poly(U)
were shown to form a triple-stranded structure
(at neutral pH, in the presence of MgCl,) with
the stoichiometry 1A:2U where the second
poly(U) strand was bound via Hoogsteen pairing.480482Studies
using poly(C) and poly(G) or
poly(1)showed triplex formation at weakly acidic
pH values (with one homopurine and twohomopyrirnidine
strand^).^^^-^^^ Later, the possibility
of formation of poly(dG).poly(dG).poly(dC) at
neutral pH was demon~trated.~It was shown'
that base triplets T-A-T and C-G-C+ can be
formed in both polyribo- and polydeoxyribonucleotide
series.487492Triplex structures were observed
not only in homopolymers, but also in
copolymers with alternating sequence containing
all purines or all pyrimidines in each strand; for
example, poly(dTdC).poly(dA-dG).poly(dT-dC)
was formed below pH 6 in the presenceof MgC1,.
Recently it was shown that these polynucleotides
can adopt triple-strandedstructure even at neutral
pH if the cytosine residues are me~hylated.~~~It
was proposed that in the triplex the third strand
was associated with a duplex through Hoogsteen
pairing in the major groove with parallel orientation
to the homopurine stranď.485.492.493Polynucleotides
with mixed purine-pyrimidine sequences
such as, for example, poly(dAdT).poly(dA-dT),
did not form triplex structures.
Guanosine and its derivatives are known
to form tetrameric aggregates at high concentrat
i o n ~ . ~ ~ ~ ~Poly(G) and poly(1) form quadruple
he lice^.^"'-^^^ The ability to form four-stranded
structures was observed also in guanine-rich polypurine
copolymers.504
DNA and RNA multistrandedstructures have
long been considered mainly an interesting property
of some synthetic polymers with no biological
relevance. Recent discoveries of supercoilstabilized
triplexstructures in plasmid DNAs and
the occurrenceof quadruplex structuresin G-rich
,b.telomericsequ.enca"(seeSection V1.D.1) turned
the multistranded structures into the latest "hits"
of DNA structure r e s e a r ~ h . ~ ~ ~ . ~ ~ ~ "
2. lntramolecular Triplexes
The discovery of Z-DNA in synthetic polyand
oligonucleotides by means of physical techniques
(see Sections I, II, and V1.B) induceďa
search for Z-formingnucleotidesequencesin various
genomes and studies of relations between --.
their location, structure, and biological function.
The history of the discovery of the supercoilstabilized
triplex DNA followed a different path.
First, a great number of papers were published
in the first half of the 1980s describing the location
and properties of the polypurine.polypy;.imidine
(polypu-py) sequences (reviewed
in Reference 506; see Section VTI.C).
Such sequences were found in a variety of eu-.
karyotic organisms and tissues located mainly
within the regulatory regions of active genes,
Many of these sequences displayed a marked hy-
,
persensitivity to the single-strand selective nucl-S
1,,suggesting the presenceof non-B DNA
structure.To explain this hypersensitivityseveral
models were postulated (reviewed in Reference
328), including slipped s t r u c t ~ r e ~ ~ - ~left-handed
non-Z DNA,509a structure in which AaT WatsonCrick
pairs alternate with Hoogsteen G-C pairs
(with G in syn conforrnati~n),~'~and a heteronomous
structure with a dinucleotide repeat
a. H-DNA Model
In 1985 Lyarnichev et al.,'" using 2-D gel
electrophoresis, observed a sharp structural transition
within the insert (from the histone gene
unit of sea urchin) of the recombinant plasmid
pEJ4containing(A-G),, in the polypurinestrand.
This transition depended on pH and on the degree
of negative supercoiling, while the mobility drop
was pH-independent (within the given pH range),
consistent with the presence of nonintenvound
complementary strand throughout the (dAdG);(dT-dC),
stretch but not with cruciform or
ZDNA. On the basis of their results they suggested
a new structure called H form, stabilized
by hydrogen ions and including a hair~in."~Considering
the results of Lee et aL5I0obtained with
poly (dT-dC)-poly(dG-dA)at weakly acid pH they
proposed a protonated triplex H-DNA. Stemming
from different methodological approaches Christophe
et al.5" suggested a protonated triplex
structure. In the H-DNA proposed by
Lyarniche~~'~the Watson-Crick duplex extends
to the center of (dT-dC);(dG-dA), tract (Figure
5) and the second half of the homopyrimidine
strand folds back upon itself, winding down in
the major groove of the helix. This returning
strand forms Hoogsteen pairs with the purines of
the Watson-Crickduplex cytosines being protonated.
The second half of the homopurine strand
also folds back, but is probably unstructured.
Neither the experimental data of Chrisfophe.
et al."' based on mapping of the nuclease S1
hypersensitive sites nor those of Lyarnichev
et al.175.5'2represented sufficient proof of the
H-DNA triplex. Such proof was obtained 2
years later by means of chemical
247,264.265.290.313.326.513
6. Chemical Probing: Strong Evidence for
H-DNA Model
In 1987 we probed supercoiled plasmid p a 4
(constructed in the laboratory of M. Frank-Kamenetskii)
using Os,py as a probe for the homopyrimidine
strand and glyoxal for the homopurine
strand in combination with nuclease S1 to
cleave DNA at the site of the modified bases.264
We demonstrated the dependence of the site-specific
modification on pH, NaCl concentration (this
dependence was observed at pH 6, bul not at pH
4), and supercoiling. At pH 5.6 a major sitespecific
modification was observed in the middle
of the homopyrimidine strand and a minor site
close to the end of the tract. On the basis of these
results we concluded that under the given conditions,
protonated triplex H-DNA is present in
the supercoiled pEJ4 DNA. At pH 4.0 similar
site-specific modification was observed even in
linear DNA m o l e c ~ l e s . ~ ~ ' ~ ~ ~Employing basically
the same approach several laboratories, including
ours, published in 1988 modification patterns
at single-nucleotide resolution (using
chemical cleavage of DNA or terminationof transcription
at modified bases instead of nuclease
S1) of several recombinant plasmids containing
(d~-d~),.(d~-de),tracts. These patterns displayed
strong Os,py modification in the middle
of the homopyrimidine strand (Figure 5) (corresponding
to the hairpin loop in H-DNA) and a
minor modification of the 3'-end of the (dT-dC),
tract (consistent with the reactionat the B-H junction),
plus a strong DEPC modificationof the 5'half
of the homopurine strand corresponding to
the 5'-half of the unpaired sequence not included
in the triplex: This hypersensitivity of the 5'-half
was observed regardless of the insert orientation
within the Other single-strandselective
chemical probes such as hydroxylamine and
methoxylarnine yielded more or less the same
modification patterns, showing that the results
are independent of the nature of the chemical
p r ~ b e . ~ ~ ~ . " ~Modification with DMS showed that
guanines of the 3'-half of the purine strand are
protected against alkylation in agreement with
their assumed involvement in the Hoogsteen pairing.215-313-326-"3The
interconversion between duplex
and triplex occurred within a few minutes,
as detected by nuclease P1 site-specific cleav-
FIGURE5. Structure of H-DNA. (A) Schematic representation
of H-DNA in (TC-AG),,, with 3'-half of the pyrimidine (dT-dC)
repeat donated to the triplex, forming the H-y3 conformer. The
5'-half of this repeat,plus the complementary 3'-half of the (dAdG),,
polyurine repeat, act as the acceptor helix in this corforrnation.The
two halves of thepolypyrimidinestrandin the tr~plex
(- and ----) are antiparallel. Watson-Crick base pairs
areshownaslines, Hoogsteenbasepairsasdots. (B) Hydrogenbondingschemes
for base tripletsin H-DNA. (C) Alternativeuse
of 3'- or 5'-half of the (dT-dC) repeat as the donated strand, to
form H-y3 or H-y5 conformers of H-DNA. Watson-Crick base
pairs are shown as lines and Hoogsteen base pairs between
the acceptor purines and uncharged T or protonated C+pyrimidines
are shown by and +, respectively. Boxed letters and
lettersin circles indicate nucleotides that are reactive to singlestrand
selective probes (e.g., Os,bipy or DEPC),whenthe DNA
is in the H conformation.
-.
A 0s.PY
Pur
age.s14The energy parameters of the B-H transition
in supercoiled DNA were obtained by
Lyamichev et al.sls for (CIA-dG), and (G), sequences.
The energy of nucleation of the H'form t
in (A-G), sequences F, = 18 kcal/mol was close
- to the corresponding value for the
Becauseof theseresults, Lyamichevet al.51sconsidered
the possibility of H-form extrusion in
homopu*pytracts shorter than 15 bp as very improbable.
This is in agreement with the recent
B
C G C
results of K o h ~ i , ~ ~ ~who demonstrated triplex
formation in (G16).(C16)but not in (G14).(Cl,).
Triplex formation in both (C-T), and (C), sequences
effectively protected the DNA duplex
h m UV-induced pyrimidine dimerizati~n.~The
degree of protection depended on the degree of
supercoiling and acidity.516
Displacement of the single-stranded region,
a characteristic feature of H-DNA, was questioned
by some authors.328Htun and DahlbergS1'
R
T A T
C
s ' - C T A C C ~ C G G T C m m T t T m C T
~ ~ - ~ ~ ~ c ~ ~ c ~ c c f f i f f i ~ ~ f f i f f i f f i f f iH-y3
TCTCTCTCTCTCTCTC
5 ' 3 '
C- C
A- T
I IC - C
T- A C- G
T- A A - T
A- T A - T
T- A A - T
T- A T - A
T- A A-T
G- C A - T
C- C A T - A
I I G- C
3 ' 5' 5 '-GTXGAGCGG T
* A - T
C * C - C
TOA-T
FIGURE 5C
demonstratedan ability of the (dT-dC),, sequence
cloned into circular MI3 phage DNA to form a
complex with H-DNA in supercoiled plasmid
containing (dT-dC),,-(dA-dG),, sequence. The
complex was studied by electron microscopy, 2D
gel electrophoresis, and chemical probing. The
latter technique showed Os,py reactivity in the
pyrimidinestrand similar to that of uncomplexed
supercoiled DNA; the DEPC reactivity of the
purine strand was significantly reduced, however,
suggesting the single-strandedness of the
5'-half of the (A-G),, stretch in H-DNA, a result
not consistent with other hypothetical models.328
A similar approach has been applied to reveal
the presence of intramolecular triplexes in (dAdG)stretches
in the initiation region of a dehydrofolate
reductase replicon of Chinese hamster
cells.fi8Formation of the complex with the (TC),
stretch in M13 DNA requires the presence
of triplex structure, in contrast to complexes between
short (T-C), oligonucleotides and (A-G).(TG)
sequences in duplex DNA.363-5'7
The results summarized above (1) provided
strong evidence in favor of the intramolecular
. triplex H-DNA and (2) suggested that from the
two possible isomers of H-DNA (Figure 5), only
that in which the donated strand comes from the
3'-half of the pyrimidine strand prevailed under
the given conditions.
c. Nucleotide Sequence Requirements
The homopu-py nature of the sequence suitable
for the formation of H-DNA is determined
by the necessity of forming base triads TAT and
CGC' or alternatively ATA and GCG (see below).
The sequence must contain a mirror repeat
(i.e., to be the same in the 3'-5' as well as in
the 5'-3' directions along a single strand), which
is also termed H-palindrome todistinguishit from
normal palindromes (inverted repeats) extruding
the cruciform structure.506.514.519-521H-palindrome
may contain a nonpalirldrornic segment (which
may even disturb the homopu-pycharacter of the
sequence) in its center.519In triplexes with identical
12-base triads in the stem, 4 to 10 base
interruptions inpthem'nter were tolerated.s21
Longer triplex loops, however, required higher
supercoil energy for triplex formation and were
less thermostable than triplexes with shorter loops.
For example, the thermal stability of the triplex
with 10-base loop was lower by 3 to 4°C and the
free energy was about 5 kcal/mol higher than in
the most stable triplex with a 4-base loop. A 12base
nonhomopurine spacer between two (dAdG),
segments did not prevent the formation of
H-DNA (stable over a broad pH range), while
the presence of a 46-base spacer with a random
sequence completely abolished the H-DNA formation
under the same condition^.^^^.^^^ If, however,
the spacer was composed of a regular (dAdT),
tract, a complex H-DNA with a cruciform
structure formed by the spacer.
The presence of bases disturbing the H-palindrome
in its noncentral region was not so well
t ~ l e r a t e d . ~ ' ~ - ~ ~ ~Mirkin et constructed
sequences
5 ' - A A G G G A G A A X G G C G T A T A G G C G Y A A 3 '
in which X and Y were either A or G. When
X = Y this sequence (inserted in pUC19 DNA)
exhibited facile transition to H-DNA. For X #
Y this transition was much more difficult or im-'
possible, as detected by 2-D gel electrophoresis.
A more detailed picture was obtained with Os,py
and DEPC modifications. For example, a 42-bp
pwpy sequence with three consecutive interruptions
formed a mixture of two smaller triplexes
instead of a large triplex containing unpaired bases
in the stem. A shorter sequence with a single
interruption formed a triplex with one unpaired
base in the stem. The presence of this interruption
resulted in a decrease of thermostability by 7°C
and the requirement of a higher energy of supercoiling
for the triplex formation. Lower requirements
for supercoiling and increased thermostability
were observed with increasing G+C
contentof a homopwpy sequence, suggesting that
the presence of CGC triads is very important for
triplex formation. No triplex formation was observed
in (dA)20*(dT)20even at superhelix densities
more negative than -0.09.5'3 It was shown
in 1990 that much longer (T,)-(A,,) segments
adopted the triplex form.524
d. Hinged H-DNA
Htun and Dahlberg247proposed a three-dimensional
model for H-DNA, from which they
predicted that H-DNA would introduce a severe
kink in DNA molecules. Studying the electrophoretic
mobility of DNA fragments containing
H-DNA in (A-G), sequence (at pH 4) at different
sites of the molecule, they showed that such a
kink does exist and that it possesses a limited
flexibility; thus, it can also be termed a hinge.
The presence of a kink or a bend in supercoiled
pEJ4 plasmid containing H-DNA was demonstrated
at pH 5.6 by electron m i c r o s c ~ p y . ~ ~
' ,
e. Conformersof H-DNA
Htun and Dahlbergs17-s26proposed a standard
nomenclature for describing H-DNA structures.
According to this nomenclature, H-DNA refers
to conformations with triple-stranded and displaced
single-stranded regions (not considering
the nature of the base pairing). To distinguish
between two possible conformers, H-DNAs in
which the donated polypyrirnidine strands come
from the 5'-half were termed H-y5 and those
from the 3'-half H-y3. The number of base triplets
within H-DNA can be indicated in brackets,
e.g., H-y5(16). Ln H-DNAs where the polypurine
strand is donated, y (for pyrimidines) is replaced
by r (for purines), e.g., H-r5.
Once formed, H-DNA may be trapped in a
local energy minimum; the initial nucleation event
can thus determine which conformer can be
formed. To form the nucleation site, rotation of
the helices flanking the donated pyrimidinesmust
occur. Formation of H-y3 conformer from (dTdC),,.(dA-dG),,
requires the relaxation of about
one more negative supercoil than does formation
of H-y5 due to the necessity of additional rotation
prior to nucleation. To form H-y5 conformer, the
donor duplex only folds back on the acceptor,
while H-y3 nucleation requires one complete turn
of the two strands around each other. Thus, Hy3
nucleation would be promoted at higher levels
of negative supercoiling when compared with nucleation
of H-y5. DEPC modification of topoisomers
containing :H-DNA confirmed this ass
~ m p t i o n , ~ ~ ~whiche& be used to explain the
reported bias toward utilizing of the 3'-half of
the polypyrimidine strand as a donor.247*265J13-513
Relative energetics costs of H-DNA formation
decrease with increasing length of the
(dT-dC).(dA-dG) repeats.526Longer repeats support
generation of multiple conformers and relax
a larger amount of negative supercoiling upon
H-DNA f o r m a t i ~ n . ~ ' ~ . ~ ~ ~Multiple forms may
result from the long repeats in which the mutation
did not occur at the repeat center.s26The - - .
pPP1 plasmid containing (dAdG),ATCGATATATATCG(dA-dG),
sequence,
which is not very long by itself but contains a
long spacer region, displayed at pH 4.5 DEPC
modification patterns consistent with the presence
of both conformers at a wide range of -u
values (0.02 to 0.1).522 With shorter purpyr sequences
of about 25 to 30 bp, only the H-y3
conformer was observed at pH 5 in a wide range
of superhelix d e n s i t i e ~ . " ~ . ~ ~ ~Similarly, the
(CAA),TCC(GAA), sequence yielded only H-
y3 conformer at a from -0.001 to -0.09 at pH
6 and 7.s26 On the other hand, increasing the
length of the regular (A-G), sequence resulted in
the formation of multiple conformers.526bTheir
presence was manifested by complex modification
patterns of (G-A),,, (G-A),,,. and (AGGAG),,
segments at pH 5.5 and below at moderate
and high superhelix densities. It thus appears
that the mechanism suggested by Htun and
DahlbergSz6might be limited to certain sequences,
while in other purpyr sequences different
mechanisms may come into play and give
" rise to different triplex structures.526a
f. Effect of pH, Ions, Supercoiling,
Nucleotide Sequence, and Length of the
Homopupy Tracts
Low pH, negative supercoiling, and increasing
length of the homopu-py sequences act interdependently
to stabilize H-DNA.B7.264.433-526b.S27
With increasing lengths of the homopu-py tract
(at pH 5.2), less negative superhelix densities
were required to induce the triplex format
i ~ n . ~ ~ ~ . ~ ~ ~AS the pH was increased (at a constant
insert length), more supercoiling was necessary
to drive the B-H t r a n s i t i ~ n . ~ ~ ~ . ~ ~ ~ ~
The homopu-py sequences capable of adopting
H-form triplexes can be composed either of
mixed sequences forming H-palindromes or of
homopolynucleotidestretches such as (G);(C),.
Mixed HomopwpySequences: With highly
supercoiled DNA modification patterns characteristicfor
H-DNA were observed even at neutral
pH.247On the other hand, in plasmids (containing
segments of regions flanking the C2b and C2a
immunoglobulin constant region genes) where
(dC-dT), tracts were adjacent to alternating pu-pyr
segments, increasing the superhelix density did
not result in triplex formation at neutral pH.528
This was explained by the ability of the 2-DNA
flanking sequence to prevent triplex formation.
The absence of triplex in pyrimidine sequences
at pH 8 at high superhelix dens'ity,:combined with
the formation of left-handed DNA in an alternating
purine pyrimidine region located 76 bp to
the 5'-side of the CT segment, was observed by
Johr~ston.~'~
Using nuclease PI,site-specific cleavage was
observed starting from the mean superhelix density
of about -0.04 at pH 4.6 and about -0.06
at pH 7.5 with both synthetic (dTdC)12.(dAdG),,
and (dG),,-(dC) Similar results were obtained
using Os,py rnodificati~n.~~'At pH >7.8
the reactivity of the homopyrimidine strand was
but the homopurine strand remained
reactive to DEPC even above pH 9. This result
was explained by the presence of a new conformation
(J-DNA)247and alternatively by different
conditions under which DEPC and Os,py reactions
were conducted.S26bIt should not be difficult
to solve this question experimentally using different
chemical probes.
Evans and Efstratiadi~~~'showed that nuclease
S1 cleavage of (dG-A);(dC-dT), sequences
is length dependent; they observed hypersensitivityof
(dG-A),,.(dC-dT),, sequences to
venom phosphodiesterase from Crotalus adamanteus
at pH 9 (this enzyme possesses an intrinsic
endonuclease single-strand selective activity).
Recently it was shown526bby Os,py and
DEPC probing that increasing the length of the
insert decreases the dependence on acid pH for
triplex formation. The (dG-dA),, sequence (constructed
by Evans and Efstratiadi~~~')adopted a
triplex structure at neutral pH and a moderate
level of supercoiling (o= -0.049). The pK of
theotriadC"GC is between 7 and 8; thus, at more
alkaline pH values unprotonated CGC triads may
contribute to triplex ~ t a b i l i t y . ~ ' ~ . ~ ~ ~
An interestingobservation was recently made
by Bernues et al. ,530.530awho showed that the (GA-C-T),,
sequence cloned in SV40 can adopt an
altered structure at neutral pH in the presence of
ZnZ
+ ions (at millimolar concentrations). On the
basis of Os,py and DEPC modification patterns
the authors proposed a triplex structure denoted
as *H-DNA stabilized by GGC and AAT triads
(Table 5). GGC and AAU triads were shown to
be stable at neutral pH.53'-532At pH 4.5 the
d(GA-CT),, stretch assumed the usual H - f ~ r r n . ~ ~ ~
It thus appears that this sequence can assume two
different structures, depending on pH and ions
present in solution. *H-DNA was observed also
in the presence of Mn2
+ and CoZ+ at both neutral
and acid pH.526bThe switch region of IgA immunoglobulin
in mice cloned into a recombinant
plasmid showed hypersensitivity in the (AGGAG)?,direct
repeat to nuclease S1 (at acid pH) -
TABLE 5 nucleases was observed in a long AT-rich poEffectof
pH and Ions on Formation of ,,
lypu.py tract (found downstream of the chicken
Protonated and Unprotonated H-DNAa myosin heavy chain gene) at pH 4.5 and 7.5.s33
Recent CD studies of poly(dA-dG)*(poly(dT-dC)
Form: PY'PY'PU PU'PU.PY demonstrated six different conformational
- states of this polymer at pH values between 8.0
F p y ppy
and 2.5.5" These results suggest that local structures
other than H triplexes may be formed in
purpyr sequences at various pH values.
(dG);(dC), Sequences: After Lyamichev et
demonstrated the existence of protonated H
P" , triplex in (G);(C), sequences by 2-D gel elecSequence:
trophoresis, Kohwi and Kohwi-Shigemat~u~~~
(A-G)ns(C-T)n showed that the (dG),,.(dC),, sequence can adopt
PH Acid Neutral another triplex structure at neutral pH in the presMe2+
No Zn2
+ Mn2
+ ence of Mg2
'ions (at millimolar concentrations)
Co2
+
formed by unprotonated base triads GGC (Table
(G)n.(C)n
PH Acid Neutral
5). (dG),,.(dC),, tract within the 5'-flanking reM
e + No Mg2
+Mn2
+Ca2
+ gion of the adult chicken PA globin gene dis--
(A);(T), In n = 69, but not in
n = 33
PH Neutral Not observed
Me=+ Mg2
+
See text for details.
and P1 (at neutral pH) at a more negative than
-0.002.s27The nucleaseS1 hypersensitivity was
retained for the shorter repeat (AGGAG),, which
displayed at acid pH Os,py and DEPC modification
patterns characteristic for the H-y3 conformer
of H-DNA.
Long rnirrorrepeats (GAA),TCC(GAA), and
(GGA),TCC(GGA), form H triplexes at pH 6.0
and 7.0 in plasmids at -a as isolated from E.
~ o l i . ~ ~ ~ 'With an increase of -a and/or lowering
of the pH, Os,pyand DEPC modification patterns
were observed that were inconsistent with the
formation of a usual H triplex. It has been suggested
that a structure is formed that simultaneously
contains two triplexes in the given sequence.
One possibility is that Hy-5 forms in the
(GAA), and Hy-3 in the (GAA), segment, but
other triplex models are possible. Similarly,
(GGA),TCC(GAA),, which is not a mirror repeat,
formed non-B conformation at acid pHs.
The energy required for the transition was about
the same as with the H triplex formed by the Hpalindrome
sequence, but the Os,py and DEPC
modificationpatternsof the former structure were
not consistent with H triplex DNA. A non-B
played similar properties. In addition to Mg2
+
ions, Ca2
+ and Mn2
+ also induced the (G),,.(C),,
sequence into a GGC triplex. Recent results suggest
that in forming the dG.dG.dC triplex structure
the potential of poly(dG).poly(dC) depends
on the length of the (dG).(dC) tract.s3s ZnZ+,
Cu2
+,and Co2
+ ions did not induce triplexstructure
in the (G),,-(C),, sequence, but did influence
thepstructureof tQe- , direct repeat sequence
adjacent to the sequence.
d[(G)z4C(G),,I~d[(C)2,G(C)2,1in pG46C plasmid
displayed structural transition at neutral pH in the
presence of MgZ
+ at ALk - 15 accompanied by
a release of five turns as detected by 2-D gel
electrophoresis.537
A 64-bp GC-rich polypu.py tract from the rat
long interspersed DNA element displayed two
classes of supercoil-dependent reactivity toward
chemical probes.538One class consisted of highly
sensitive bases (whose sensitivity was strongly
affected
by pH and Mg2
+ ions) probably contained
in H-DNA triplexes observed earlier in
(dG);(dC), sequence^.^*^^^^^^^^ The other class
comprised moderately sensitive bases independent
of reaction conditions. This reactivity suggests
the presence of non-B-DNA, but the structural
basis for this reactivity is not known.
(A);(T), Sequences: FoxSz4 showed that
(dT),,.(dA),, sequences adopt an intramolecular
triplex structure in supercoiled plasmid (at native
a)in the presence of Mg2
+ at pH 8. The structure
structure hypersensitiveto ;ingle-strand selective displayed a characteristic Oslpy modification in
the centerof the (T,) segment and a DEPC modification
of the purine strand charac'ťeristicfor the
H-y3 conformer. No chemical modification resembling
the triplex structure was observed in
(A3,).(T3,) and (A2,)*(T2,) segments. The nuclease
S1 cleavage and DEPC modification patterns
of (A,)*(T,) suggested that the displaced
half of the purine strand might weakly interact
with the triplex. 2-D gel electrophoresis failed to
detect any structural change; this was said to be
due to the relatively small fraction of molecules
containing the triplex. Compared with (Gn).(Cn)
and (A-G);(T-C), sequences, formation of triplex
in (A,,)*(TJ segments requires much longer
homopuspy stretches. suggested that the
requirements for long (A).(T) tracts may be due
to the high stability and rigidity of propellertwisted
(A,)*(T,) helix containing bifurcated hydrogen
bonds (see Section m).If this explanation
is correct, why did (A15T15)segments show facile
C-type cruciform extrusion253and why did thermally
more stable (G,).(C,) sequences with n <
69 form triple^?^*.^'^."^ Further work is necessary
to better understand recently discovered triplex
structure in long (A).(T) segments.
3. lnfermolecular Triplexes
a. Complexes of Oligonucleotides with DNA
The possibility of complexing oligonucleotides
with DNA has been long a n t i ~ i p a t e d . ~ ~ ~ . ~ ~ ~
Formation of intermolecular triplexes resulting
from the interaction of oligonucleotides with the
complementary homopurine sequence in duplex
DNA was demonstrated only recently.3s359.363.540540
Interest in intramolecular triplexes has recently
greatly intensified (see Section V.D.2.e), mainly
due to their potential application as inhibitors of
gene expression and as recognition and cleaving
elements in chromosome mapping.
It has been shown that a 27-base-long
oligonucleotide probe binds to duplex DNA
at a single site within the 5'-end of the human
c-myc gene, forming a colinear triplex with the
duplex binding site.s* The triplex formation correlates
with repression of c-myc transcription in
vitro. A shorter 11-mer oligopyrimidine
d(TITCCTCCTCT) formed a complex with a
fragment of double-stranded DNA containing a
complementary The complex
was stabilized by additional binding energy resulting
from intercalation of the aromatic ring
system (acr-dine,phenanthroline) attached to the
oligonucleotide 5'-end. The oligomer was bound
to the major groove of DNA in a parallel orientation
with respect to the purine strand. This
kind of highly sequence-specific interaction can
be used to control DNA expression.
A similar approach has been adopted to develop
a new strategy of genomic DNA mapping.359Oligonucleotides
with attached cleaving
agent EDTASIron(I1) at the 5'-end can induce
sequence-specific double-strand breaks in DNA.
An oligonucleotide of about 15 to 19 nucleotides
should be sufficiently long to recognizea specific
sequence in the human genome,356.359providing
in a formal sense lo6times better resolution than
natural restriction nucleases. More details can be
found in recent reviews and also in Section
V.D.2.e.356359.541
b. Physical Studies of Triplexes
Ln addition to the studies mentioned above,
structural information on the triple helices has
been deriveti' from poly- and oligonucleotide
studies by means of physical techniques. Early
X-ray fiber diffraction studies of homopolynucleotides
resulted in a triplex model in which the
third strand lies in the major groove, interacting
with the duplex via Hoogsteen pairing.493-"2The
duplex portion of the structure adopts an A-like
conformation with a C3'-endo sugar pucker. the
third strand can be homopurine or homopyrimidine,
depending on experimental conditions. If
the third strand is homopyrirnidine it is parallel
to the homopurine strand in the triplex.
Recent studies of oligonucleotides with various
nucleotide sequences by means of NMR and
other techniques confirmed in the principle triplex
model, resulting from the X-ray fiber diffraction
m e a s ~ r e m e n t s . ~ ~ ~ . ~ ~ - ~It has been shown
by NMR, CD, and other measurements that at
neutral pH in the presence of MgCl,, triple-helical
(dA),,.2(dT),, is formed.546In this triplex
thymine N3-H iminoprotons are involved in both
Watson-Crick and Hwgsteen base pairing.
d(GA), and d(TC), octarners are able to form
B-DNA duplex as well as triplexes dependent on
-
11
experimental condition^.^^^.^^^ pyrpu-pyr triplex
was observed at low pH and in an excess of
pyrimidine strand. The results unambiguously
conf'ed that the second pyrimidine strand binds
via a Hoogsteen pairing (with cytosines protonated
at N3) in the major groove of a Watson-.
Crick duplex. The conformation of the pyrimidine
sugars was A-DNA-like (see Section 11),
while purines had a B-DNA-type sugar pucker.
Under the given conditions the TA Hoogsteen
base pairs appeared more stable than GC+
Hoogsteen base pairs. Results of independent 2D
NMR studies of 11-mers were in good agreement
with those of octarner studies showing Ahelical
base stacking conformation in the oligopurine
strand of the 11-mer Quite recently,
Sklenar and FeigonW8constructed a 28base
DNA oligomer with a sequence that could
potentially form a triplex containing C'GC and
TAT triads. Their 2-D NMR experiment showed
that this oligonucleotide forms an intramolecular
triplex at pH 5.5 and that a significant amount
of triplex remains at neutral pH.
The 26-mer ~ ( G A A G G A G G A G A ~ ~ ~ . C T C C T C C I T C )
formed a hairpin in solution.s49If this oligonucleotide
was mixed with d(TCTCCTCCTTC) at
pH 5, a triplex is formed as detected by gel electrophoresis
and CD measurements. The triplestranded
structure melted in a biphasic profile.
The duplex-to-triplex transition was accompanied
by an average change in enthalpy of -73
( 2 5 ) kcal/mol. The equimolar mixture of
d(CTCTTCTTTCTTTTCTTTCTTCTC) and
d(GAGAAGAAAGA) formed a triplex at pH 5
as detected by gel electrophoresis, ethidium bromide
interactions, DNaseI digestion, and CD
spectra. Cooperative thermal transition was observed
that was attributed to the disruption of HDNA-like
structure into single strands
X-ray fiber diffraction studies of
poly(dG)*poly(dC) complexes with N-aacetyl-L-arginineethylamide
displayed a B-form
pattern, although these polynucleotides favor (in
the absenceof arginine) the A-DNA form.5" Upon
dehydration a triplex was observed, most likely
formed by poly(dC+)-poly(dG)*poly(C).
the Frank-Kamenetskii group is now well establ
i ~ h e d . " ~ . ~ ~ ~The structure can be formed in any ?
homopu-pysequence containing a mirror repeat.
Strong evidence for the protonated triplex strut`ture
(Figure 5) has been supplied by chemical*
probing.ZM.265.290.313.326.513Important structuralinformation
has been extracted from recent NMR
studies on synthetic oligonucleotide^.^*^ In spite
of a large amount of data, it is still difficult to
ascertain the structure of the half of the strand
not involved in the triplex base pairing. Is it truly
single-stranded and free of any interactions?This
does not seem very probable. At least some results
of chemical probing suggest that bases in
this part of H-DNA in supercoiled plasmids are
involved in interactions that are most probably
substantially weaker than those within the trip
l e ~ . ~ ~ ~ . ~ ~ ~ - ~ ~ ~ - ~ ~ ~DO these interactions involve the
triplex or are they limited to, for example, stacking
interactions within the displaced single strand
itself? Ln the former case we would have to consider
the possibility of the formation of a loosely
bound tetraplex.522
D. Other Structural Changes
, The structures of cruciform, left-handed
DNA, and triplexes,'as well as their relations to
DNA supercoiling and other factors, were well
established in the 1980s. These structures are
undoubtedly not the only ones stabilized by negative
supercoiling. Recent experimental data suggest
that a number of other structural changes
may ~ c c u ~ . ~ ~ - ~ ~ ~Presently, some data cannot be
interpreted in terms of a well-defined structure,
while others do not reveal the relation of a structural
change to DNA supercoiling. It is to be
expected
that these points will be clarified and
new local DNA structures discovered soon.
Among the DNA local structural changes that
have recently attracted the greatest attention are
base-unpaired regions, structures of telomericsequences,
and parallel-stranded DNA. These are
briefly discussed in the following paragraphs.
1. Structure of Telomeric Sequences
4. Conclusions
The triplex H-DNA structure suggested by
Telomeres (reviewed in References 505 and
556 to 558); the ends of eukaryotic chromo-
somes, all have a similar type of npcleotide se:
quence,i.e., tandemly repeated GC-rich sequences
with a pronounced strand-specific base
composition asymmetry such as d(C4AJ-d(T,G?)
(repeated at least 50 times) in ciliated protozoan
Tetrahymena or d(C4A4)d(T4G4)of Oxytricha.
Telomeres are involved in the replication and
maintenance of the chromosomal ends, and a
higher order structural theme shared by all telomeric
sequences has been suggested to be critical
for their function.559In the last 2 to 3 years
evidence has accumulated showing specific local
structures in telomeric sequences.
a. C,A Hairpin
Budarf and B l a c k b ~ r n ~ ~ ~demonstrated that
the telomeric sequence poly d(C4A,);d(T,G4), of
Tetrahymena inserted in a plasrnid is hypersen:
sitive to nuclease S1under conditions of negative
supercoiling. No such hypersensitivity was observed
in linear DNA. As in the case of homopuapy
sequences, mapping of nuclease S 1 hypersensitive
sites did not yield a basis sufficient
to deduce the higher structure of the given sequence.
Using 2-D gel electrophoresis and chem;
ical probing, Lyarnichev et a1.561.56'a proposed a
novel protonated DNA conformation, the C,A
hairpin. In this structure two independent hairpins
are formed that are stabilized by C.C+ and
A.A + base pairs.
The model of (C,A) hairpin structure is based
primarily on the results of chemical probing.
DMS, Os,py, and KMnO, showed no protection
of G's and Ťs from modification, suggestingthat
the G4T2strand is virtually unstructured. In the
C4A2strand the strongest reaction was observed
in the central A's, while other adenine residues
were lessreactive. The nonequivalenceof the two
strands in the (C,A) hairpin is supported by the
binding of the oligonucleotide (C3A,C4C)complementary
to the G-rich strand, which occurred
only at acid pH. (G4T2),cOmp~emedtGto the '
C-rich strand showed no binding.
The data presented by Lyamichev may also
be consistent with triplex H-DNA carrying CGA
and ATC triads (in addition to TAT and CGC'),
if two conformers are present in comparable
Further studies, including modification
of C in various topoisomers, may help to
solve this problem.
b. Quadruplexes
In the last few years studies of sysheticoigonuclep_tidescontaining
telomeric sequence motifs
have resulted in new models of DNA structures
based on non-Wat~on~Crick-- base-pairing.
In 1987 Hendersonet al.559investigated the structure
of G-4c_h strqnds of several telomeri_c_s_ecp?n!es
&d demonstrated the format& of novel
intramolecular structures containing G-G pairs
with guanosine.-. - residues in syn conformation as
determined by NMR spectroscopy. Sen-- -.--and
Gilbert562probed the structure of self-associated
si-ngleIstranded DNA oligonucleotides (which
displayed a decreased electrophoretic mobility
when compared with single strands) by DMS and
concluded that four-strandee
d-- structures were
formed in which th-e strands runjnparallel fashion
and guanines &-bonded to eaih other by
~ o o ~ steen
Quite recently compelling evidence was
gathered independently from three laboratories
.showing that G-rich oligonucleotides may form
anti~arallelqhhdruplexescontaining cyclic guanine
base tetrad^.$^^-^^' Sundquist and K . l ~ g ' ~ ~prepared
a series of oligonucleotides containing duplex
sections with a FGGGG repeating sequence
(found in Tetrahymena telomeres) followed by a
single-stranded 3'-terminal overhang of two repeats
and showed that these oligonucleotides dimerize
to form stable complexes in solution. The
complexes were u~liauelvs t a b w d bv mtassium
ion, The dimerization was mediated by the 3'terminal
overhang. In these complexes the N7 of
every guanine in the 3'-tenninal overhangwas
inaccessible to DMS and DEPC, while both pairs
of thymines were accessible to osmiumLe-trp~ige.
The authors proposed that telhkefc DNA dimerizes
by hydrogen bonding between two intramolecular
hairpin loops, forming antiparallel
quadruplexes.
In contrast to Sundquist and Williamson
et d.$@used single-stranded oligonucleotides
composed of the telorneric sequencerepeats
from h t r i c h o and ~etrah~rneha,while
Panyutin et al.565used fragments of DNA con-
taining (dG),,, (dG),,, and (dG),,, respectively.
The resultsobtained by different methods (chemical
pr~bing,'~~-'~~UV cr~sslinking,'~electrophoretic
m ~ b i l i t y ~ ~ ~ . ~ ~ ~ " ~ )induced proposals of
remarkably similar structures with guanosine
bonds of adjacent chains in opposite conformations:
syn vs. anti. Thus, in any tetrad two guariosine
residues have anti and two have syn conformations.
The suggestedss9~s62-s65Hoogsteen
base pairing (Figure 5) received further support
from experiments with replacement of dG with
dl at various positions of the telomeric seq~ence.?~The
oligonucleotide complexes prepared
by Sundquist and opposed those
of Williamson et al.'@ by the different effect of
monovalent ion on their stabilization, suggesting
differences in sizes of cavities created within the
structures.
Do these results mean that the original model
of a parallel-strandedquadruplex (G4-DNA;Figure
6) proposed by Sen and Gilbert562should be
abandoned? The recent work of these authors
suggests that this should not be the case.567Sen
and Gilberts6' showed that both ggi#e&tranded
G4-DNA (with all of its guanosine residues $'
the g& conformation) and ytiparal1eJ 'quadruplexes
can be formed. They expected sequences
with four or more separated runs of G's to yield
four-stranded intramolecular fold-back structures,
sequences with two runs to produce dimer
fold-back structures, and sequenies with single
runs to yield G4-DNA. Complex sequences
formed parallel-stranded G4-DNA in the pres-,
ence of sodium and rubidium-butnot the presence
of potassiumions. This anomaly(which was
not found in oligonucleotides containing short,
single runs of three or more Gs) arose because
potassium ions stabilized quadr~~lexstructures
SQ strondy that transient intermediateswere tragped,
preventing formation of the G4 structure.
onn nation of G4-DNA was very strongly dependent
on the ratio of Na+/K+.Two phases of G4
formation were shown: at low molar ratios of
potassium, a progressive increase in the G4 formation
rate was observed, while at higher potassium
ratios G4 DNA formation was inhibited.
Telomere DNA sequences were observed in
phylogenetically diverse organisms including
protozoa, fungi, and even higher eukary~tes."~
It now seems that telomeres of all eukaryotic
nuclearchromosomes may have basically similar
structures that can undergo transitions in dependence
on their ionic environment. We can, however,
only speculate about the relations between
theirspatial organizationand biological function.
2. Base Unpaired Regions
Early melting was perhaps the first recognized
structural change assisted by the free energy
of negative s~percoiling.'~~-~~~'Local base
pair disruption was ~ b s e r v e d ' ~ ~ . ~ ~ ~under various
conditions far from melting but differing from
physiologicalconditions. Hypersensitivityto single-strandselective
nucleases(such as mung bean
and P1 nucleases) was reported at physiological
conditions in AT-rich regionsof supercoiledplasrnids.223.568.569The
hypersensitive sites did not
correlate strictly with AT content and appeared
in replication origins and transcriptional regulatory
regions in both pro- and eukaryotic DNAs.
Until recently it was not known whetherthis local
DNA hypersensitivity was due to permanent or
transient unwinding of the given DNA region.
Recent experiments with 2-D gel electrophoresis
showed that the unwound structure is thennodynamically
stable (prevailing at equilibrium over
the B-form).!70-57'
Transition to stsble unwinding occurred"%t
about -a = 0.05 and the average extent of
unwinding was sufficient to completely unwind
the region recognized by the nuclease at -a
0.067.570,e formation of completely unwound
DNA segments in replication origins and transcriptional
regulatory regions is a very attractive
suggestion, since DNA must unwind to initiate
replicationand transcription. The evidence based
on nuclease hypersensitivityand 2-D gel electrophoresis,
however, is not sufficient to rule out
other alternatives.s69-571
Kohwi-Shigemats~~~~showed that the yeast
replication origin sequence is reactive to
CAA;s69~572.s73this result suggests that the reactive
region is truly unpaired. It also has been
shown that stable unpaired regions hypersensitive
to CAA modification surround the immunoglobulin
heavy chain (IgH) enhancer. These regions
are AT-rich and contain negative regulatory elements.
Unpaired CAA-hypersensitive, AT-rich
FIGURE 6. (a) Scheme for the formation of G,-DNA. The formation
of the dimer structure G,must be rate-limiting, and it must
be rapidly converted into G,. Only three possible structures (including
K and G;)of fold-back intermediatesare shown, but other
st~cturesmay also be formed. (b) Structures of product K most
compatiblewith its methylation-protectionpattern. The methylationenhancedguanine
is circled. The arrows indicate the 5'-3' direction
of the sugar-phosphate backbone. (Reprinted by permission from
Sen, D. and Gilbert, W., Nature, 244, 410, 1990.)
',
regions
are also included within matrix association
regions. Similar to mung bean nuclease hypersensitive
sites of the E. coli replication oegin,n1the
structural transition in the IgH enhancer
region occurs at a about -0.05, and increasing
the negative superhelix density results in extension
of the unpaired region.573A local high AT
content is not sufficient to cause the DNA segment
to adopt the unpaired structure. Mutation
of three adenines (to either G or C) in the sequence
ATATAT in the CAA-hypersensitive region
resulted in a marked reduction of its sensitivity
to CAA. It was proposedthat the ATATAT
mdtif may be kinked, serving as a nucleation site
for base unpairing.
Various sequences required different environmental
conditions to form unpaired region~;"@"~for
example, at 22OC the 3'-sequence
of the IgH enhancer was very sensitive to CAA,
while the autonomously replication sequence of
the yeast origin was unreacti~e,'~~showing some
CAA hypersensitivity at 37°C.573Base unpairing
is involved in the C-type cruciformextrusion (see
Section VI.A.2). Hypersensitivity to single-strand
chemical probes under superhelical stress was.
observed also in other AT-rich sequences3
' and
the curvature-inducing ATATAITITTTAGAGATITIT
sequence. 'I3
The coincidence of the base unpaired regions
with various functional elements such as negative
regulatoryelements573(repressingIgHenhanceractivity
in fibroblasts but not in Pcells), nuclear rnatrix
association regions, replication origins,s7G572
and transcriptional regulatory regions suggests
that DNA structure may play an active role in
replication and transcriptionand other biological
functions rather than serving only as a passive
source of nucleotide sequence information.
3. Parallel-Stranded DNA
Multistranded structures containing parallel
strands were discussed in preceding paragraphs.
Recently it has been shown that parallel-stranded
duplexes can be formed by oligonucleotidesconsisting
of AT pairs under physiological,-condiParallel-stranded
DNA was formed
in hairpin molecules (1) with stems stabilized in
a parallel orientation by 5'-5' or 3'-3' phosphodiester
linkages in the hairpin loop574.577-579and
(2) with crosslinked ends.S80-582
From the biological point of view the most
interesting parallel-stranded DNA appears to be
that formed by hybridization of two complementary
oligonucleotides with appropriate sequences
(partially homooligomeric A-T sequences).
Such parallel-strandedmolecules were
recently designed, synthetized, and characteri
~ e d . ~ ~ ~ - ~ ~ ~ . ~ ~ ~ ~ ~ ~The conformational constraints
in these oligonucleotides was obtained by using
overhangs or by appropriate combination of block
and mixed sequence^."^ Runs of alternating (AT),
segments579as well as interspersed GC pairs
are compatible with the parallel-stranded helix
but induce its destabili~ation.'~
a. Structural Features of Parallel-Stranded
DNA
The structure of parallel-stranded DNA has
been studied in solution using various physical
and chemical techniques. (Crystallographic data
are not yet available). The results of gel electre
phoresis, W absorption, and CD spectra measurements,
as well as thermal meltingand chemical
probing, suggested that parallel-stranded DNA
contains a distinct secondary structure.57k587Differences
between antiparallel- and parallelstranded
DNA behavior were observed upon
binding intercalating drugs chemical probing and
nuclease cleavage.574-578.583-585.587
According to the theoretical force field calculati~n,'~~reversed
(trans) Watson-Crick AT
base pairs are formed in parallel-stranded DNA
in which the adenine 6-amino group is hydrogen
bonded to thymine via the 0, instead of the 0,
of the keto group [in the normal (cis) WatsonCrick
pair]. The reverse (Watson-Crickbase pairing)
has been supported by 'H-NMR578DMS
probing,574and Raman spectra.579The Raman
NOESY rneasurement~,~'~and molecular
m ~ d e l i n g ' ~ ~ . ~ ~ ~suggested that the furanose
rings are mainly in C2'-endo conformation and
bases h e in anti oI5Cntation. 31P-NMRof the
intramolecularparallel-stranded hairpin (containing
(A)-(T)stem and four cytosines in the loop)
and a 25 bp parallel-stranded duplex composed
of two complementary strands [with a (A),,~(T),,
block and sequencescontaining TA and AT steps]
produced different result^."^^^^^ In the former
parallel-stranded hairpin no drastic differences
between parallel-and antiparallel-stranded DNAs
were observed,578whereasthe backbonestructure
of the latter parallel-stranded duplex differed from
that of the antiparallel-stranded duplex.587No clear - explanation
of these results has been yet offered.
Measurements of fluorescence resonanceenergy
transfer of 5'-fluorescence labeled oligonucleotides
confirmed that the strands in parallelstranded
DNA indeed have the same p~larity.'~
Parallel-stranded DNA inserted in a supercoiled
plasmid underwent a transition to a (pu),.(pyr)
triplex as detected by 2-D gel electrophoresis and
chemical probing.591The data confirmed that parallel-stranded
DNA forms a right-handed duplex.
Parallel-stranded duplexes formed between unnatural
a-anomers and natural poligonucleotides36s~5mas
well as duplexes containingT*Tpairs
and phosphate-methylatedbackbone
have been reported.593
6.Biological Role
Does parallel-stranded DNA occur in vivo,
and, if so, what is its biological role? This question'cannot
yet be answered, but the stability of
parallel-stranded DNA at physiological conditions
suggests that such a structure might be
formed in vivo. It has been suggested that parallel-stranded
DNA would arise most readily in
the course of exchange reactions involving interactions
between separated segments of DNA,
including recombination and genomic rearrangem
e n t ~ . ~ ~ ~ . ~ ~ ~In principle, parallel-stranded DNA
can be formed in several ways, the simplest of
which involves intrastrand loop formation.576
Parallel complementary sequences were found
in the Drosophila genome593aand the ability of
very similar sequences to form parallel-stranded
DNA in vitro was demon~trated.~~~.Thus,the
possibility of the existence of local parallelstranded
DNA regions in vivo does not seem to
be unrealistic. More details concerning parallelstranded
DNA structure and properties can be
found in recent review^.^^^.^^^
E. Conclusions
It has been shown that negative supercoiling
stabilizes left-handed DNA segments, cruciform,
triplex and C,A-hairpin structures, and base unpaired
regions. These structures are frequently
called "unusual structures". Are these structures
unusual only to us, because we recognized them
much laterthan B- and A-DNA, or are they really
unusual in nature? Does, for example, %DNA
or the triplex structure occur in vivo substantially
less frequentlythan A-DNA? Until now we have
not been able to answer such questions.
A common feature of cruciforms, triplexes,
C,A hairpins, and base unpaired regions is the
accessibility of some bases for interactions with
the environment (in contrast to bases hidden inside
the B- or A-DNA double helices). The aqcessibilityof
bases in 2-DNA is less marked than
in other unusual structures, but the results of
chemical modification experiments suggest that
at least in (A-C).(G-T) sequences left-handed
DNA segments (in supercoiled plasmids) may
contain a number of exposed bases.467470In addition,
accessible bases were found at B-Z junctions
regardless of the nucleotide sequence of ZDNA.
Local accessibility of a small fraction of
bases was observed earlier even in nonsupercoiled
DNAs far below melting conditions.lo Regions
containing accessible bases have been called
"open"DNA regions. Similarly, cruciforms, triplexes,
hairpins, and B-Z junctions may be called
open DNA structures. It is not clear whether the
base unpaired regions are completely structureless
(this does not seem to be probable) or whether
they adopt some structure. In the latter case they
can be included among open DNA structures.
If the open DNA structures play some biological
role, it would be surprising if the accessibility
of their bases would not take part in biological
processes. Exposed bases in open DNA
structures may represent targets for specific interactionswith
other DNA and RNA strands;they
c& facilitate or @event recognition of a given
sequence by specific proteins and other substances
interacting with DNA. Many mutagenic
chemicals would react preferentially with exposed
bases in a way similar to that of chemical
probes. In addition to open structures discussed
above other open structures and complexes can
be transiently formed during various biological
processes, including DNA replication, transcription,
and recombination. An example of such
structures is an open complex of RNA polymerase
with promoter sequences. This complex contains
a so-called melted region 12 to 16 bases in
length. Recently DEPC, DMS, and Os,py have
been applied to characterize this region in an open
complex of the lac UV 5 p r o m ~ t e r . " ~ . ~ ~ ~It has
been shown that bases in the templatestrand from
-10 to +3 react with DEPC (A residues) and
DMS (C residues).259Thymines at positions -8,
-9, and - 11 reacted with Os,py in contrast to
those at +1 and +2 (assumed to be in a single-
stranded region) that did not react. On the other
hand, thymine at -11 showed hypersensitivity
toward Os,py even though this base is expected
to be outside the locally "melted" region. These
results suggest that the locally melted region is
not completely structureless. Its detailed ch&acterization
by means of chemical probes may
provide important information, including those
about the strength of the given promoter. Work
with this goal in mind is under~ay."~Similar
approaches can be applied to study other open
structuresand complexesboth in vitro and in situ.
VII. DNA STRUCTURE IN THE CELL
Much information has been gained concerning
DNA structure in solution, fibers, and crystals,
although information about DNA structure
in its natural environment, i.e., in the cells, is
scarce. Such information is, however, vital for
a better understanding of the biological role of
DNA. The preceding sections showed that DNA
supercoiling stabilizes various local structures
such as cruciform, triplexes, left-handed DNA,
etc., and that extrusion of these structures in turn
changes the DNA superhelix density. Extrusion
and absorption of the local structure may represent
an efficient way of regulation of the biological
processesinvolvingDNA, but do these structures
really exist in vivo?
Early experiments suggested that a certain
portion of DNA isolated from various organisms
had a single-stranded character(reviewed in Reference
10). It was shown that the amount of DNA
cleaved with single-strand selective nuclease S1
increased during the period of DNA synthesis in
human diploid fibroblasts.s95Ln experiments carried
out with isolated DNA it was difficult to
exclude the possibility that regions with a singlestranded
character were formed secondarily (e.g.,
due to nuclease cleavage a.fter the cell disruption).
Techniques have become available that
make possible the study of DNA structure inside
the cell.
A. Methods of Analysis of DNA
Structure in the Cell
Both indirect and direct ways of demonstrating
the formation of open local structures in the
cells have been employed. The former include:
(1) methods of studying certain genetic consequences
of the formation of such a structure,
including the modulation of transcription by ZDNAs%and
cruciform,s97and the susceptibility
of potential Z-forming
sg8
and inverted repeat
sequence^^^^-^' to deletions: and (2) methods
based on changes in DNA topology induced by
the formation of a local s t r u ~ t u r e ~ ~ ~ . ~ ~ ~and on
compensation of the simultaneous partial DNA
relaxation by intracellular topoisomerases (linking
number assay). The indirect methods do not
significantly disturb cell life, but interpretation
of the results may not be free of ambiguity; for
instance, the linking number assay of the cruciform
extrusion in vivo relies upon the accurate
regulation of superhelix density in the cell. It is
expected that the change in the superhelixdensity
induced by the structural transition will be compensated
for by the regulatory system, which reduces
the DNA linking number so that the level
of supercoiling is ree~tablished*~~Unfortunately,
other changes in DNA superhelix density (e.g.,
due to a specific protein binding to the inverted
repeat) may be compensatedfor in a similar way.
Perhaps the most efficient way to demonstrate'the
formatioq.~fa local DNA structure in
the cell is by probing DNA in situ. This has been
achieved by molecular gene ti^^^.^^ and chemical
approaches.39 The molecular genetic method induces
an enzyme in the cell and theconsequences
of its specific interaction with the local DNA
structure in the cell &e determined. For example,
cleavage of the cruciform loop by 'I7 endonuclease
was used to demonstrate cruciform structure
in E.. coli cells.603Single-strand selective
chemical probes have been applied successfully .
in DNA structure studies in vitro (Section V).
We have showd9that Os,bipy enters E. coli cells
without disturbing their integrity and site-specifically
reacts with bases in open regions of the
cellularDNA such as B-Z junction^^^-^ or triplex
structures.604Testing of other osmium tetroxide
complexes revealed that in addition to Os,bipy,
Os,TEMED (Figure 2) can be applied to probe
DNA structure in s i t ~ . ~ O ~KMnO, was recently
introduced as another single-strandselective probe
of the DNA structure in E. coli cells.4'Chemicals
reacting with double-stranded DNA such as
DMS3m.33'.606and copper-I, 10-phenanthr~line~"~
were used to obtain footprints of intracellular
DNAs. DNA footprinting can also be carried out
by irradiating cells with short exposures of UV
light.372.608-609With the use of UV irradiation it
may also be possible to obtain informationabout
local DNA structures in vivo.
After the treatment of cells with a chemical
probe (or UV light), DNA is usually isolated and
the probe reaction sites determined as in experiments
in vitro (Figure 3). It is also possible to
detect the probe binding in the cell by means of
'irnmunofluorescence techniques without isolatingDNA.6'0By
using Os,bipy it has been shown
- - - that,Z-DNA,39,40*248.447cru~iform,~~-~"and triplex
structure^^^.^ can exist in E. coli cells.
B. Cruciform Structures
Inverted repeats, the potential cruciform sequences,
arefrequently found within genetic regulatory
regions.260p612.613Can all of these sequences
extrude cruciforms in vivo? Reports
published in 1983 s ~ g g e s t e d ~ ~ ~ . ~ ~ ~that in bacterial
cells cruciform may be rare because of its slow
formation. On the other hand, the tendency of
some inverted repeats to undergo deletion was
~ b ~ e r v e d , ~ ' ~ . ~ ~ ~ . ~ ~ ~ - ~ ~ ~suggesting- the possible
presence of cruciforms in the cell.
I. Evidence of Cruciform Structure in
Bacterial Cells
Using the linking number assay, Hanniford
and P~lleyblank~~*concluded that the (A-T),, sequence
extruded cruciform in E. coli cells under
conditionsof blocked protein synthesis. In 1987,
Panayotatosand Fontainem3showed cleavage of
the ColEl inverted repeat in the intracellulqr
pLAT75 plasmid by T7endonuclease induced in
E. coli cells. The digestion site coincided with
the cruciforh loop cleavage by T7 and S1 nu:
clease in vitro. The ColEl inverted repeat in the
pLAT75 was' placed in its native environment .
within the coding sequence of the colicin resistance
gene and was actively transcribed from its
natural promoter. The sequence surrounding the
palindrome was highly AT-rich. Thus, the combined
effects of active transcripti~n'~~and presenceof
adjacentAT-rich sequences3
' (seeSection
V1.A) could significantly contribute to cruciform
extrusion in the cell. By means of the Os,bipy,
probe, the presence of cruciform in E. coli cells
was demonstrated recently in (A-T), inserts of
different length^.^.^" In the system not undergqing
active transcription, biased in favor of cruciform
formation (using salt shock or topoisomerase
mutation to increase the superhelix
density) cruciform was detected in (A- T),, (AT),,,
and (A-T),, but not in (A-T),,. These experiments
made it possible to calculate the effective
DNA superhelix density inside the cell
that responded directly to genetic and environmental
influences.
The above data seem to be sufficient to warrant
the conclusion that at least some inverted
repeats can form cruciform in E. coli cell^.^^^.^"
The presence of AT-rich C-type-inducing sequence
and superhelix density increased above
the average intracellular level appears to be important
for cruciform extrusion in the cell.
2. Biological Role
If we admit that cruciforms can exist in the
cell, a question concerning their biological funcZion
may arise.::It has been suggested that one of
the signals in;dlved in the initiation of DNA synthesis
is a specific local DNA structure that is
recognized by the enzyme necessary for replic
a ~ i o n . ~ ' ~ . ~ ' ~Features common to many r-e tion
origin sequences include inverted reueatsand
AT-rich region^.^'^^^'^ Thus, cruciform structures
may be one of the candidates that might be involved
in si nalinginitiati~n-ofofDNAsynthesis.
Quite recentlyg__;__War et a1.616observed somecorrelation
between the distribution of activated
origins of replication and the distribution of cruciform
using irnrnunofluorescenceof eukaryotic
nuclei labeled with monoclonal anticrucifonnantibody.
By meansof fluorescenceflow cytometry
they determined the number of cruciforms to be
around lo5per nuclei. It was shown that human
and rat Ells contan cruciform-binding protein
that is structure-specific and sequence-independ:
ent.617.6'8This protein was identified as nuclear
HMGl .6'9 HMGl is an abundant component of
the nucleus involved in transcriptions and DNA
replication; its interaction with the crucifo&
structure points to an important biological role
of this structure.
It has been shown that a synthetic E. coli
promoter containing a cruciform in a -10 region
may regulate transcription in a supercoil-dependent
Transcription from this promoter
in vitro was repressed as the cruciform [50 bp
highly (88%) AT-rich inverted repeat] was extruded.
Transcription in vivo was induced as supercoiling
was relaxed due to DNA gyrase iAhibition
[which was less (64%) AT-rich]. A
cruciform in the -35 promoter region behaved
similarly in ~itro;~~Ohowever, the same inverted
repeat had littleeffect on the transcription in vivo.
A 48-bp inverted repeat placed in the J-F intercistronic
region of $ x 174 replicative form DNA
gave identical transcripts in vitro with extruded
and unextruded cruciform.621Unusual local
structure, most probably cruciform, prevented
both transcriptional initiation and elongation in
form V of pBR322 DNA.622
It has been shown that transcription induces
the generation of positive supercoils in front of
the transcription complex and formation of negative
supercoils behind it.14s-623-625Thus, C N C ~ ~
form structures can be absorbed or extruded dying
transcription. S m extrusions in vivo are
ratherjmprobae, while the occurenceof$&=
extrusion In vivo, especially in highly AT-rich
inverted repeats, appears probable.-Therefore, the
difference observed in the effect of inverted repeats
contained in the -10 and -35 regions on
transcription in vivo may be due to different extrusion
pathways of the two cru~iforms.~~'In vitro
transcription proceeds in a less complex milieu
(e.g., no membrane attachment sites are
present), which may result in the formation of
supercoil domains differing from those created
in vivo by their size, degree of supercoiling, etc.
It is thus not surprising that the effect of an inverted
repeat on the transcription in vitro and in
vivo is not always the same.
The recent data suggest that cruciform structures
can exist in the cell and might be involved
in the regulation of DNA replication and tranr
rip ti on.^^^^^^^^^^^^^^^^^^ The possibility of in-,
volvement in other biological processes cannot
be excluded.626.627
C. Left-Handed DNA
The presence of 2-DNA in fixed eukaryotic
chromosomes was reported in the early
1980s.628-630By means of anti-2-DNA antibodies
patterns of different intensity bands in polytene
chromosomes of Drosophila and Chironomus
were observed by indirect immunofluoroesce
n ~ e . ~ ~ ~ 'Shortly after this occurred it was sho&
that 2-DNA could not be detected in unfixed
chromosomes isolated by micromanipulaand
that solvents used in fixation procedures
induced different stable DNA tracts, resulting
in the reproducible patterns. Therefore,
antibody binding results can only be taken as
evidence for the presence of potential 2-DNA
sequences in the eukaryotic genomes. Attempts
to demonstrate the existence of 2-DNA in (GC);(G-C),
segments of plasmid DNA in E. coli
by the linking number assay635did not solve the
problem of the existence of 2-DNA in vivo since
the results of this assay cannot be interpreted
unambiguously (see above).
1. Existence of Left-Handed DNA in
Prokaryotic Cells
, Tn 1987 ,we used Os,bipy to search for lefthanded
DNA in (c-G);(c-G), segments of an
intracellularp l a ~ r n i d . ~ ~ - ~We showed that Os,bipy
reacts at the boundaries of these segments, recognizing
site-specifically the B-Z junction inside
the E. coli cell. At the same time Jaworski
et using a special molecular genetic technique,
showed independently that left-handed
DNA exists in the plasmid (C-G);(C-G), stretches
and elicits a biological response in E. coli cells.
Their in vivo assay was based on the in virro - .
observation that a EcoRI recognitionsite was not
methylated when it was near or in the 2-helix.
In the in vivo assay a plasmid encoding the gene
for a temperature-sensitive EcoRI methylase
(MEcoRI) was cotransformed with any one of
several plasmids containing (C-G), or (T-G),
blocks of differentlengths with target EcoRI sites
in the centeror at the end of the blocks. Inhibition
of methylation by the MEcoRI was observed for
the inserts, with the longest (G-C), segments long
enough to form 56 bp left-handed helices. These
results provided evidence that left-handed DNA
can exist in the living cell; however, the pos'sibility
that the observed inhibition of methylation
was due to a specific protein binding cannot completely
be excluded. On the other hand, the site- .
specificOs,bipy modification of the B-Z junction
in the cell could not be due to protein binding,
although this assay did not allow the study of the
biological response of left-handed DNA.40.139
Considering the results of the genetic42 and
chemical35studies performed in 1987, it may be
concluded that strong evidence has been obtained
showing that left-handed DNA can exist in the
cell. This conclusion is supported by further studies
of the inhibition of MEcoRI in vivo and a
thorough linking number assay602.636using systems
similar to those used by Jaworski et al.42as
wellas by the insert deletion a n a l y s i ~ . ~ ~ ~ . ~ ~ ~Using
the linking number assay and MEcoRI inhibition
assay, Zacharias et al.636showed that cytosine
methylation stabilized left-handed DNA in E. coli
cells just as occurred in the in vitro experiments.
Insert deletion analysis showed that sequences
capableof adopting Z-DNA were generallystable
when cloned into an untranslated site of pBR322
(EcoRI), but suffered deletions when cloned into
a site (BarnHI) located in the tetracycline resistance
structural gene.598
Using Os,bipy, Rahmouni and Wells248recently
showed that (C-G) segments as short as
12 bp adopted left-handed DNA when cloned
upstream from the tet gene, whereas no lefthanded
DNA was found when the(C-G)blocks
(upto 74 bp long) were cloned downstream. These
studies strongly support the notion of
with varying degrees of supercoiling in E. coli
cells, perhaps related to transcriptional activity.
The important contribution of R. D. Wells' laboratory
to studies of Z-DNA in bacterial cells
was reviewed r e ~ e n t l y , ~ ~ . ~ ~The progress made
in the last few years allows us to conclude that
left-handed DNA can exist in bacterial cells [at
least in (G-C), segments], and that its existence
is dependent on the level of intracellular '
supercoiling.
2. Left-Handed DNA in Eukaryofic Cells
Studies of Z-DNA in eukaryotic cells by
means of antibodies are more difficult ,because
of the possibility of the secondary induction of
Z-DNA by (1) direct effect of fixatives, (2) removal
of proteins during fixation procedures,
which results in changes in DNA superhelix density,
and (3) perturbance of the B-Z equilibrium
by the anti-Z-DNA antibody. To overcome these
difficulties, Wittig et al.638.639 encapsulated permeabilized
myeloma nuclei (that were active in
transcription and replication) in agarose microbeads
and probed the extent of Z-DNA formation
in dependence on supercoiling. Upon binding ZDNA
specific antibody, they observed a broad
plateau of constant binding that was taken as a
measure of preexistingZ-DNA in the nuclei. They
calculated that about 0.04% of the base pairs
were in the Z conformation.639Inhibition of topoisomerase
I with camptothecin resulted in a
higher Z-DNA content, while cleavage with
DNaseI induced the complete loss of preexisting
Z-DNA in the nuclei.
Soyer-Gobillard et al.639alocalized Z-DNA
in limited areas inside the chromosomes of the
dinoflagellate Prorocentrum micanr by irnrnunocytochemistry
on squashed fmed, unfixed, and
frozen cells. This organismus is a primitive eukaryote
whosechromosomes show a permanently
well-organized DNA skcture with no histones
and a nucleosomal system that would modulate
DNA supercoiling. This makes the dinoflagellate
chromosome a highly suitable model for studying
Z-DNA and other local structures in vivo.Z-DNA
was localized often at the periphery or near the
segregation fork of dividing chromosomes. In the
nucleolus, Z-DNA was observed only in the nucleolus
organizer region and never in the fibrillogranular
area. Positive results obtained with
unfixed and frozen cells provide strong evidence
for the existence of Z-DNA in eukaryotic cells.
These results cannot be simply correlated with
the previous data obtained from eukaryotic chromosomes
because of substantial differences in
their chromosome organization.
(C-G), and (C-A), sequences were cloned
into SV40.620+641(C-G)nwas highly unstablecompared
with that of (C-A)n.620This instability,
however, was not related to the formation of ZDNA
in eukaryotic cells, as no signs of this structure
in (C-G), and (C-A),,.(T-G),, inserts in vivo
were detected by the linking number assay.641It
has been pointed out in this article that the &sults
of this assay cannot be unambiguously interpreted;
the possibility that changes induced by
Z-DNA formation were compensated for by, forexample,
dissociation of some protein molecule,
cannot be excluded.
Studies based on the dinoflagellate model introduced
by S~yer-Gobbilard~~~"as well as by the
technique developed by Wittig et a1.638.639represent
a substantial advance in the study of ZDNA
in eukaryotic cells. Further techniquesare,
however, needed to elucidate the question of the
presence, distribution, and biological function of
left-handed DNA and other local structures in
eukaryotic cells. Application of chemical probes
and antibodies specific to DNA-probe adduct?
may representa new and useful approach to these
studies. An even more interesting approach can
be seen in a novel confocal Raman microspectr~metry;~~,by
means of this technique the Raman
spectra of a single intact cell, a chromosome,
or a polytene chromosome band and
interbandcan be obtained, providing information
about DNA structure and DNNprotein ratio. This
technique brings new perspectives for future DNA
structure research in vivo.
3. Biological Role of 2-DNA
While the possibility of the existence of lefthanded
DNA in vivo has been reliably establ
i ~ h e d , ~ ~ . ~ ~the biological role of the structure remains
unclear. Long (dC-dG), tracts are not
widely found in biological systems, but (dTdG);(dA-dC),
sequenceswere found in both proand
eukaryotic genomes (reviewed in References
34 and 643 to 647) associated with many genes
(reviewed in References34, 36, and 648 to 653).
Proteins that bind to Z-DNA in vitro were isolated
(reviewed in References 34, 36, and 648 to
653). Up to the present time, however, no ZDNA-dependent
in vivo functions of these pro:
teins have been identified.
Among biological roles suggested for ZDNA
are its participation in transcription, recombination,
chromatin structure, e t ~ . ~ ~ . ~ ~ ~ . ~ ~ . ~ ~ ~It was
shown that E. coli RNA polymerase could transcribe
through (C-G),, sequencein the B-form.'"
When this sequence was flipped to the left-handed
form, the RNA polymerase together with its nascent
transcript was blocked at the boundary of
the (C-G),, tract. In vivo, however, the entire
sequence was transcribed, suggesting some
mechanism that removesor prevents the blocking
of transcriptional elongation in the cell. It was
recently shown that HMGl protein removes in.
vitro the transcriptional block caused by the (CG),,
sequence in the left-handed form.656
Insertion of (C-G), in the lac Z gene of E.
coli inhibited expression of P-galactosidase in
v i v ~ . ~ ~ 'The (C-G),, showed a 34-fold decrease
of P-galactosidase synthesis when inserted instead
of a lac operator and a 24-fold decrease
when inserted between codons 5 and 6 of the lac
Z gene. With shorter (C-G), sequences the decrease
was substantially smaller. It might be interestingto
try to compare the observed inhibition
with the actual presence of left-handed DNA in
the cell.
In contrast to (C-G),,, which caused a transcriptional
block in negatively supercoiled plasmid
in vitro, the (C-A),,*(T-G),, sequence induced
no strong hindrance to tran~cription.~"
Longer (C-A);(T-G), sequences (with n = 60
and 179 bp, located upstream of the rat prolactin
gene) formed lgft-han~DNA and inhibited gene
transcription in v i t r ~ . ~ ~ *Nucleosome assembly at
the (G-C),, insert was prevented when the insert
adopted Z-DNA in a supercoiled ~ l a s m i d . ~ ~ ~
These results suggest that left-handed DNA
may be involved in .important biological processes.
Further work is, however, necessary to
elucidate the role of this DNA form in transcription,
recombination, and other biological
processes.A
D. Triplexes
1. Occurrence of Polypugy Tracts
The occurrence of polypu.py sequences in
genomes of various organisms has been studied
intensively in the past few years. (G-A);(T-C),
tractsconstitute 0.4,0.3, and 0.4%, respectively,
of the rat, hamster, and mouse genomes, but only
0.7 and 0.5% of the human and monkey gen- !
~ r n e . ~ 'These tracts were also found in other
~ r g a n i s m s . ~ ~(T);(A), and (G);(C), sequences
are present in the human genome with 0.3 and
0.0002% frequencies, respectively.663In mice and
rats the transcription units are flanked by (A);(T),
and (G);(C), sequences.664Tracts of (C), and
(G), blocks have been found in the vicinity of
mouse immunoglobin light chain genesM5
and
within the 5'-flanking region of the adult chicken
PAglobin gene.- 0,sequences are found 5'
upstream of the human A-globin gene intermingled
with (T-G),.= (T-C), and (T-G), blocks
are adjacent in the third intron of the apoliproprotein
CLI gene.667In Drosophila chromosomes
(T-C), and (T-G), blocks show a nonrandom distribution,
with the highest occurrence on the X
chromosome.668-670 Blocks containing (GAA),
sequence units are present on all human chromosomes
except the Y chromosome.671The
(GGA)? sequence family may be ubiquitous ig
thegenomesof higher eukaryotes. Pentamers T,,
(T),C, TTClT, TTTCT, and TCTlT are represented
with over 125 copies in the ovalbumin
gene (CCT), sequences occur at recombination
sites of the complex satellite DNA of
the Bermuda land crab.673homopu.py sequences
are also found in numerous virus g e n ~ m e s . ~ ~ ~ . ~ ~ ~
Homopu*pyregions 1 1 0 bp in length were
found in human P-globin region and six other
human genes with an average 6f about one string
per 170 to 250 bp.675A high bias in favor of
homopu-pystrings in a human genome may affect
nucleosomestability and placement.M3.675-677Nucleosomes
were reconstituted with all homopwhomopyrsequences
tested,675-678with the exceptionof
poly(dA).poly(dT), which was not able
to form nucleosomes.679-680
Further details on the occurrence of homopu-homopyrsequences
can be found in recent
r e v i e w ~ . ~ ~ * ~ ~ O
2. Existence of H-DNA In Vivo
In contrast to cruciform and left-handed ZDNA,
whose supercoil-stabilized structures were
uncovered in the early 1980s, H-DNA structu&
was proposed only a few years ago, therefore,
attempts to identify this structure in vivo have
only a short history. In 1987 Lee et a1.206generated
monoclonal antibody specific to triplex
DNA. They reported binding of this antibody to
mouse metaphase chromosomes and interphase
nuclei. In fixed mouse and human chromosomes
a positive correlation between immunofluorescent
staining patterns, G- andlor C-banding patterns,
and Hoechst 33258 banding was observed.681Unfixed
isolated mouse chromosomes
were only weakly fluorescent. These results are
interesting, but they do not represent unequivocal
evidence of triplex structure in the cell, as their
interpretation suffers from drawbacks similar to
those of 2-DNA immunofluorescent
~ t a i n i n g . ~ ~ ~ . ~ ~ ~
Using Os,bipy we did not observe any sign rku
of triplex formation in pEJ4 intracellular plasmid kf.
at neutral pH.38If E. coli cells were preincubated
I u.
in pH 4.5 or 5.0, a modification pattern char- ,
acteristic for H-DNA was obtained. The shift of -,
the intracellular sup&helix density to more negative
values (by cultivating the cells in media
supplemented with 0.35 M NaCI) resulted in a
stronger site-specific modification.
A more detailed analysis of the pL153 plasmid
(containing the homopu-py sequence from
pH4 not undergoing active transcription)showed
differences between the Os,bipy modification in
vitro and in sit^.^^^ In situ, more bases were modified
in the triplex loop and modification of two
thymines at the 3'-end of the (C-T),, sequence
(forming b e B-H junction of the H-3y conformer)
was weaker (Figures 5 and 7). On the
other hand, modification of three cytosinesat the
5'-end of the (C-T),, sequence was observed;
these cytosines were unmodified in vitro. The
modification of these cytosines can be expected
in H-y5 conformer, which is formed at lower
superhelix density than H-y3 observed mainly in
isolated plasrnids. It was therefore tentatively
suggested that H-y5 conformer prevails in the
cell. Differences in triplex loop modification,
which might be due to some intracellular interactions,
suggest that the triplex structure in the
cell may differ in some details from that observed
in vitro.
Basically the same modification patterns
(Figure 7) were found in the pH range 4.5 to 5.2;
at pH 5.4 no characteristic triplex modification
was observed. When compared with extracellular
pH, the intracellular pHs were higher by about
0.5 as determined by the fluorescein method.
These results were obtained at pH values that are
FIGURE7. Modificationof the polypyrimidine strand of the insert
in pi153 plasmid with Os,bipy (A) in vitro and (B) in situ.- The
length of the vertical lines in the nucleotide sequence represents
the relative intensities of the bands on the sequencinggelobtained
by densitometric tracing after (A) treatment of supercoiled pL153
DNA in vitro at pH 5.0 and (8)after Os,bipy treatment of E. coli
cells harboring the pL153 plasmid (at external pH 5.0). Two conformers
of the H-DNA triplex are shown in which is indicated the
direction of the donated strand by full triangles. The arrows show
the strongest modified base in the triplex loop and different modificationof
bases at the potentialB-H junctionsconsistent with presence
H-y5 conformer in the cell and H-y3conformer in vitro.
not fully physiological for E. coli, but E. qoli pH values in different cell compartmentscan sigcan
grow at pH 5.0 and 5.2, and while a tiansfer nificantly vary.683In the cell, requirements for
of cells from pH 6.9to 4.3results in an induction protonation might be decreased by specific proof
acid shock proteins, no such induction occurs tein binding andlormore negativesuperhelixdendue
to the transfer to pH 5.682Intracellular pH sity in the given DNA domain. It may thus be
depends on the life cycle of the cells (low pH is expected that triplex structures will be detected
common to resting cells), and in eukaryotic cells at higher pH values, especially with longer po-
lypvpy tracts, which have demonstrated the ability
to form triplexes at neutral pH.
The possibility of the existence of H-DNA
in E. coli cells under physiological conditions is
supported by the results of the recent insert deletion
analysis of mixed polypu-py sequences capable
of forming H-DNA in ~ i t r o . ~ ~ 'Similar to
Z-formingsequences, the former sequences were
stable when cloned into a nontranslated region,
while inserts located in the tetracycline resistance
structural gene were deleted. Parniewski et al.523
showed that polypu*py sequences capable of
forming H-DNA in vitro are undermethylated in
vivo within the potential triplex loop when grown
in the JM dam' strain. dam undermethylation
was suppressed by the administration of chloramphenicol
to the cells. The results were explained
by H-DNA formation in vivo and protection
of the triplex loop from methylation by
interaction with a specific protein. This explanation
is very attractive, but formation of H-DNA
with the given plasmids at neutral pH requires a
u of -0.06; so far, only less negative values in
vivo have been r e p ~ r t e d . ~ ~ ~ . ~ "Other interpretationsare
possible:presenceof a partiallyextruded
H-DNA trapped specifically in vivo or formation
of a different structure which is reabsorbed due
to chlorarnphenicol treatment. Application
of chemical probes in situ might.,
help to clarify this problem. It has been shown
that H-DNA can exist in a bacterial cell at acid
intracellularpH value^.^'.^^^ Further work is necessary
to establish whether this structure does
exist under physiological conditions and in eukaryotic
cells, where the location and high occurrence
of polypu-py sequences suggest their
biological importance.
3. Structural-Functional Relations
Polypyrimidine sequences in vitro were observed
in a number of naturally occurring seq
u e n c e ~ , ~ . ~ ~ ~ . ~ ~some of which are involved in
known biological functions. Carciaogen-induced
amplificationof the integrated polyomavirus DNA . .
was arrested within'aspecific cell DNA segment
containing (G-A),,.(C-T),, tract.684Singlestranded
(T-C)n and (G-A)n tracts of various
length were cloned into M13 phage and replicated
by extension of the M13 primer to determine
whether these tracts act as stop signals for DNA
replication in ~ i t r o . ~ ' ~Specific arrest of replication
was detected around the middle of (T-C)n
and (G-A)n with n >16. In (T-C)n tracts the
arrests were more prominent at pH 6.5 to 7.5
than at pH 8.0. It was concluded that the arrests
are due to triplex formation between partially
replicated (T-C)n or (G-A)n tracts and the unreplicated
portion of these sequences. Cooney et
showed that a 27-base-long purine-rich oligonucleotide
binds to duplex DNA at a single
site within the 5'-end of the human c-myc gene
115 bp upstream from the transcriptionorigin P1.
Correlation between the triplex formation at -115
bp and repression of c-myc transcription in virro
was shown. If such a triplex formation can also
occur in vivo, it may represent an alternative
method of gene control in the cell.
Using Os,py, DEPC, and DMS probes
K i ~ i b u r g h ~ ' ~identified H-DNA in vin-o in a
mixed polypu.py sequence from a positive cisacting
transcription element of the human c-myc
gene. This sequence binds several transcription
factors. K i n n i b ~ r g h ~ ~ ~speculated that hybridization
of the RNA component of one of these
factors may represent the first step in H-DNA
formationin vivoand that H-DNA would increase
the transgriptional activity of the c-myc gene. An
unusual strucdre, probably H-DNA, was formed
by the (G),.ATT(G), sequence in a supercoiled
plasrnid containing the genome of the human immunodeficiency
virus type 1.687 This sequence is
located at the integration site of a human immunodeficiency
virus (HIV) provirus.
Kohwi-Shigemat~u~~'showed quite recently
that (G);(C), sequences enhance CAT gene
expression in eukaryotic cells. The level of enhancement
was highest for n = 28 to 30, comparable
to the polyoma enhancer. Any shorter or
longer tracts were less active. In vivo competition
assay suggested the existence of a transacting
factor that interacts with the (G);(C), sequence.
In Drosophila nuclei a protein binding to
multiple GAGA DNA sequence motifs was found.
This protein activates the transcription of Ubx
promoter in a binding-site-dependent manner.688
Further proteins from the same source bind to
regions of (C-T), in the promoters of heat shock
and histone genes.689Interaction of the purified
protein fraction with the intergenic region located
between promoters occurred on linear DNA fragments,
indicating that supercoiling was not required.
Interaction of these proteins with supercoiled
DNA in vitro was not studied.
In principle, specific protein interaction with
the polypu.py region might stabilize the duplex,
preventing triplex formation and vice versa. On
the other hand, formation of a triplex can prevent
specific protein-DNA interactions normally occumng
in+the duplex. Oligonucleotide-directed
triplex formation can inhibit recognition of the
DNA double helix by prokaryotic restriction1
modification enzymes and by eukaryotic transcription
factor at homopurine target
E. Conclusicms
Among the local DNA structures discussed
in this section, triplexes appear to be the best
candidates to play a role in gene expression. Iqtermoleculartriplexesseem
to have a betterchance,
to form in vivo, as their requirements for specific
environmental conditions and negative supercoiling
are less stringent. Moreover, they can be
formed not only between DNA molecules, but
also between DNA and RNA, including variousnucleoproteins.
Involvement of intermolecular
triplexes in biological processes in vitro such as
DNA r e p l i ~ a t i o n , ~ ~ ~ . ~ ~ ~trans~ription,~~~and restriction/m~dification~~~.~~'has
been demon-
strated.
The formationof intramolecular triplexes, 2DNA,
and cruciforms in vivo is strongly dependenton
the intracellularDNA superhelix density.
Until recently the probability of the formation of
these structures in vivo was considered to be rather
low, as the (average) effective level of (unconstrained)
supercoiling is about one half that of
purified DNA,173.5%.599.602.692
The recent discoveryof transcriptional waves
of supercoilingmakes the extrusion of local DNA
~tructu~smuch more probable.'45-623-625*693In fact,
if a suitable sequence is present in a sufficiently
negatively supercoiled DNA domain, a correspondinglocal
structure should be formed in vivo
unless its extrusion is prevented by some other
factor (e.g., protein binding, kinetic barrier). Detection
of a localstructure in vivocan be exploited
to study the level of effective supercoiling in a
given DNA d ~ m a i n . ~ ~ ~ . ~ ~'
Therefore, in addition to a question frequently
asked in recent years, "Do these (local,
unusual) structures exist in vivo?", we may now
also ask, "Why is this structure (e.g., 2-DNA)
not formed in the given nucleotide sequence in
vivo?" Asking questions is usually easier than
answering them. To answer the above questions,
further development of techniques suitable for
DNA structure studies in the cell is necessary.
We recently applied osmium tetroxide complexes
to study the presence of open local DNA structures
in eukaryotic cells by means of immunofluorescence.610As
the chemical probe can be
applied to cells and even glands prior to fixation,
the adverse effects of fixation are eliminated.
Using this technique we observed selective binding
of the osmium probe to DNA in the cells,
suggesting a wide occurrence of open local structures
in eukaryotic cells. These structures are
probably not limited to cruciforms, B-Z junctions,
and triplexes. They may include further ,
known and unknown structures and their junctions
as well as open transcription complexes and
other structures connected with DNA replication,
recombination; andl,other biological proc
e ~ s e s . ~ ~ ~ . " ~ ~ ~ " - ~ *The permanent or transient
availability of bases contained in these structures
for interaction with the environment (which is
important for their detection) may also play a
significant biological role.
VIII. PERSPECTIVES
The question ofrecognition of DNA nucleotide
sequences by specific proteins is one of the
most important problems of contemporary mo- lecular
biology. There is no doubt that the specific
proteins do not read the nucleotide sequence
as such; rather, they recognize the three-dimensional
structure of DNA. N u s s i m ~ v ~ ~ 'believes
that DNA regions may be distinguished by the
thermodynamicsor flexibility of the DNA double
helix and not by different local structures, because
they might not be trapped as such. Evidence
has been presented suggesting that different
protein motifs might interact with the DNA
grooves andlor backbone, recognizing their spe-
cific sequence-dependent spatial arrangement~.'~'In
fact, the amount of data about the
relations between nucleotide sequences and their
locations in pro- and eukaryotic genomes is much
larger than the amount of data concerning the
presence and location of DNA local structures.
The reason for this difference has been due principally
to the difficulties with obtaining the latter
data. The results summarized in this review indicate
that local DNA structures can be trapped
both in vitro and in vivo, and that sequences from
which these structures may be extruded are frequently
located in biologically significant sites
of the genomes. It would thus be rather surprising
if the local DNA structures did not represent any
signal in the DNA recognition. I believe that the
recent progress in the development of techniques
of probing the DNA structure in the cells will
soon result in a great advance in our knowledge
of the relation between local DNA structures and
their biological function. The present situation
seems to resemble that close to the end of the
1970s, when the suggested polymorphy of the
DNA double helix was rather reluctantly accepted.1°This
was at a time when the crystals of
the Z-DNA were probably already growing.
List of Abbreviations
i
Chemical Probes
0 s,py: osmium tetroxide,pyridine reagent
Os,bipy: osmium tetroxide complexed with
2,2'-bipyridine
phe: 1,lO-phenanthroline
bpds: bathophenanthroline disulfonic acid
TEMED: tetramethylethylenediamine
DEPC: diethylpyrocarbonate
BAA(CAA): bromo(ch1oro)acetaldehyde
DMS: dimethylsulfate
ENU: ethylnitrosourea
MNU:methylnitrosourea
CMC: N-cyc1ohexy1-N1-~(4-methy1morpholinium)
ethylcarbodiimide-p-toluene
Other Abbreviations
homopu-py: homopurine-homopyridine
polypu-py: polypurine-polypyrimidine
2-D: two dimensional
DEDICATION
This article is dedicated to Professor Julius
Marmur on the occasion of his 65th birthday.
ACKNOWLEDGMENTS
This paper was written during my sabbatical
leave at the Max-Planck Institute for Biophysical
Chemistry, Gottingen and supported in part by
grant 436 CSR-113/13/0 of the Deutsche Forschungsgemeinschaft
(DFG).
This review could not have been written
without the help of Dr. Thomas Jovin, to whom
I am greatly indebted for his hospitality, stimulating
discussions, advice, and critical reading of
the manuscript. I am also very grateful to Drs.
Donna Arndt-Jovin, Stephan Diekmann, Udo
Heinemann, Michel Robert-Nicoud, and other
colleagues for their valuable comments on sections
of this article. My thanks are also due to
Drs. Jim Dahlberg, Maxim Frank-Karnenetskii,
Udo Heinemann, Terumi Kohwi-Shigematsu, Jan
Klysik, David Lilley, Andrzej Stasiak, Ed Trifonov,
Robert D. Wells, and other colleagues for
allowing me to quote their unpublished data. Last,
but not least, I would like to express my deep
gratitude to Eleanor Mann and Dr. Frantisek Jelen
for their help with preparation of the
manuscript.
REFERENCES
1. Wilkins,M. H. F., Physical studiesof the molecular
structure of deoxyribose nucleic acid and nucleoprotein,
Cold Spring Harbor Symp. Quant. Biol., 21,
75, 1956.
2. Crick, F. H. C., The double helix: a personal view,
Nature, 248, 766, 1974. '
5. '~imrnerman,S. B., The three-dimensionalstructure
of DNA, Annu. Rev. Biochem., 51, 395, 1982.
4. Jovin, T. M., Recognition mechanisms of DNAspecific
enzymes, Annu. Rev. Biochem., 45, 889,
1976.
5. Palecek, E., Changes in oscillopolarographic behaviour
of deoxyribonucleicacids at temperatures below
denaturation temperature. J. Mol. Biol., 11, 839.
1965.
6. Palecek, E., Polarographic behaviour of native and
denatured deoxyribonucleic acid, J. Mol. Biol.. 20,
263. 1966.
7. Palecek, E., Deoxyribonucleic acid conformational
changes at temperatures below melting temperature,
Arch. Biochem. Biophys., 125, 142, 1968.
8. Johnson, P. H. and Laskowski, M., Sugar-unspe-