REVIEW ARTICLE

DNA structure and function

Andrew Travers1,2 and Georgi Muskhelishvili3
1 MRC Laboratory of Molecular Biology, Cambridge, UK
2 Department of Biochemistry, University of Cambridge, UK
3 Jacobs University Bremen, Germany

Keywords The proposal of a double-helical structure for DNA over 60 years ago pro-
A-DNA; B-DNA; alternative DNA structures; vided an eminently satisfying explanation for the heritability of genetic
DNA as an energy store; DNA backbone
information. But why is DNA, and not RNA, now the dominant biological
conformation; DNA elasticity; DNA
information store? We argue that, in addition to its coding function, the
information; DNA structure; DNA topology;
genome organisation ability of DNA, unlike RNA, to adopt a B-DNA structure confers advan-
tages both for information accessibility and for packaging. The information
Correspondence encoded by DNA is both digital – the precise base specifying, for example,
A. Travers, MRC Laboratory of Molecular amino acid sequences – and analogue. The latter determines the sequence-
Biology, Francis Crick Avenue, Cambridge dependent physicochemical properties of DNA, for example, its stiffness
CB2 2QH, UK
and susceptibility to strand separation. Most importantly, DNA chirality
Tel: +44 (0)1223 891921
enables the formation of supercoiling under torsional stress. We review
E-mail: [email protected]
recent evidence suggesting that DNA supercoiling, particularly that gener-
(Received 14 November 2014, revised 26 ated by DNA translocases, is a major driver of gene regulation and pat-
February 2015, accepted 21 April 2015) terns of chromosomal gene organization, and in its guise as a promoter of
DNA packaging enables DNA to act as an energy store to facilitate the
doi:10.1111/febs.13307 passage of translocating enzymes such as RNA polymerase.

Introduction
DNA is now the predominant genetic material in the bacteria. More, it confirmed Chargaff’s fundamental
living world. But which properties of this long polymer discovery of the equivalence of A and T and of G and
favour this ubiquity? Why DNA and not, say, the C bases in double-stranded DNA [3]. Most impor-
structurally very similar double-stranded RNA? tantly, the structure implied that the information in
Just over 60 years ago Watson and Crick published the DNA base sequence could possess, by virtue of the
their classic paper [1] on the structure of DNA. In it, complementarity, the ability to be replicated into two
they emphasized two principal features of the molecule identical copies. This insight into the fundamental
– the complementarity of the base sequences on the basis of genetics has underpinned the immense
two strands and the double-helical nature of the poly- advances in genetic understanding and manipulation
mer. The base sequence complementarity, with adenine over the last 60 years.
complementary to thymine and guanine complemen- Today, however, the feature of DNA that defines
tary to cytosine, provided an elegant molecular expla- the molecule is the fact that the two strands are
nation for the discovery in the previously decade by entwined as a right-handed double helix [4]. DNA is
Avery, McCarty and Macleod [2] that DNA was likely ‘the double helix’. Although this double-helical charac-
the ‘transforming principle’ that enabled the transfer ter is not required by the complementarity per se – a
of genetic information between different strains of simple straight ladder structure would fulfil this func-

Abbreviations
DNA, deoxyribonucleic acid; FACT, facilitates chromatin transcription/transactions; FIS, factor for inversion stimulation; HMGB, high mobility
protein B; H-NS, histone-like nucleoid structuring protein; RNA, ribonucleic acid.

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2279

DNA structure and function A. Travers and G. Muskhelishvili

tion just as well – it does impart crucial physical ing in the form of a coiled helical axis may affect hun-
and chemical properties to the polymer. It is these dreds of base pairs [7,8].
properties that play a major role in the biological Direct and indirect readout of DNA-recognition
function of DNA. The genetic functions of DNA can sites by proteins is a major determinant of binding
thus be understood as the synergism of two properties selectivity. In direct readout, the individual bases in a
– a tape containing the information store encoding the binding sequence make direct and specific contacts to
sequences of proteins and RNA molecules and a poly- the protein surface, whereas in indirect readout, the
mer existing as double-helical string enabling the pack- binding affinity depends on recognition of a structure,
aging, accessibility and replication of the information such as a DNA bend or bubble, whose formation is
store. Crucially, both the coding of proteins and RNA influenced by DNA sequence, but does not in general
molecules and also the physicochemical properties of require a protein contacting a specific base. In practice,
the polymer are specified by the base sequence. DNA recognition by proteins effectively spans a con-
tinuum from completely digital to completely analogue
with many proteins utilizing both modes.
DNA as an information store
For both modes of recognition, the DNA double
What is the nature of the genetic information stored in helix differs from, and is arguably more effective than,
DNA? The distinction between a linear code responsi- the RNA double helix. Direct readout requires inti-
ble for specifying the sequences of RNA and protein mate contact between exposed chemical groups on
molecules and also sequence-specific recognition by both the protein and nucleic acid surfaces. For DNA
DNA-binding proteins, and an equally important more recognition, direct readout in most examples takes the
continuous structural code, specifying the configura- form of a DNA-binding motif being inserted into the
tion and dynamics of the polymer extends the informa- major groove. In this groove, different exocyclic
tional repertoire of the molecule. Both these DNA groups of the bases in a pair are exposed compared
information types are intrinsically coupled in the pri- with those in the minor groove (Fig. 1). Consequently,
mary sequence organization, but whereas the linear although A–T and T–A base pairs in a sequence are
code is, to a first approximation, a direct digital read- distinguishable by the position of the thymine methyl
out [5,6], the structural code is determined not by indi- group charge pattern in the major groove, in the
vidual base pairs, but by the additive interactions of minor groove, the exposed charge patterns of T–A and
successive base steps. The latter code, being locally A–T base pairs are identical. Similarly, the charge pat-
more continuous, thus has an analogue form [5,6]. terns of C–G and G–C base pairs in the major groove
Importantly, the manifestation of analogue properties are distinguishable by the relative position of the 4-
is dependent on the length of the DNA sequence. For amino group of cytosine. Again, however, there is little
example, under physiological conditions, DNA difference in the relative spatial arrangements of the
unwinding manifested as melting may be restricted to charge patterns of C–G and G–C base pairs in the
a short sequence (say up to ~ 10 bp), whereas unwind- minor groove. The major groove thus provides more

Fig. 1. Exposure of chemical groups of nucleotide bases in the major and minor grooves of DNA. M, major groove; m, minor groove.
Yellow, thymine 5-methyl group; blue, basic groups: adenine 6-amino group (major groove), cytosine 4-amino group (major groove) and
guanine 2-amino group (minor groove); red, exposed cyclic nitrogen atoms and oxy- groups. Note that the presence of the thymine 5-methyl
group in place of hydrogen in uracil the enables A–T base pairs to be distinguished from T–A base pairs in the major groove. (Adapted with
permission from IMB Jena Image Library).

2280 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

sequence information than the minor groove. How- Although the formation of a B-type structure is a
ever, importantly, the wide and shallow morphology crucial aspect of DNA functionality the factors which
of the DNA major groove is in stark contrast to the shift the A M B equilibrium are, apart from water
narrow and deep structure of the RNA major groove. activity, poorly understood. One aspect is base-type.
This pattern is reversed for the minor groove. For a In principle, the coding capacity of DNA can be
protein DNA-binding motif, particularly one contain- achieved not only by the canonical A–T and G–C base
ing an a-helix, access to the DNA major groove is pairs, but also by other possibilities. For example, a
more facile than to the minor groove. This fundamen- DNA polymer with diaminopurine–thymine (DAP–T)
tal difference between DNA and RNA follows directly and hypoxanthine–cytosine (H–C) base pairs with,
from their chemical structures. Whereas DNA can respectively, three and two interbase hydrogen bonds
adopt (at least) two forms of right-handed double-heli- (Fig. 3) would, in principle, present a similar potential
cal structures, A-DNA and B-DNA (Fig. 2A), RNA for protein recognition and thermal stability [15].
can only form an A-type double helix because of the Other variations would be DNA molecules in which
steric restrictions imposed by the 20 hydroxyl residue all the base pairs contain either two or three hydrogen
on ribose [9–12]. The B-DNA structure, that proposed bonds [16]. However, not only do the component bases
by Watson and Crick [1], is most stable at high humid- specify a digital code, they also affect the physico-
ity, but converts to the A-form as the water activity is chemical properties of the molecule. For example,
lowered [13]. On this argument, it is the ability to DNA molecules with a reversed pattern of hydrogen
adopt the B-form that facilitates direct access to DNA bonding (DAP–T and H–C base pairs) more readily
sequence information. adopt an A-type conformation than DNA with the
Not only does the A ? B transition affect direct canonical base pairs [16,17]. This is because the prop-
readout, it also changes the physicochemical proper- erties of the double helix depend not only on the base-
ties of the polymer. An A-type double helix is, on pairing capacity of the constituent bases, but also on
average, stiffer than a B-type double helix and conse- the stacking interactions between adjacent base pairs.
quently distortion of A-DNA to a particular bent Changing base-pairing interactions by effectively trans-
configuration is energetically less favourable than for ferring a 2-amino group from guanine to adenine
the corresponding distortion in B-DNA [14]. Such dif- (thereby creating hypoxanthine and diaminopurine)
ferences would be expected to favour B-DNA as the changes the overall stacking because the charged 2-
preferred substrate for packaging involving tight amino group, by being in a different immediate chemi-
DNA bending. cal environment, also affects the dipole moments asso-

A B
(a)

(b)

Fig. 2. (A) Structures of A-DNA and B-DNA. Note the difference in groove width and the relative displacements of the base pairs from the
central axis. Reproduced with permission from Arnott [12]. (B) A–T and G–C base pairs shown for Watson–Crick pairing (a) and Hoogsteen
pairing (b). syn and anti indicated different sugar conformations. Reproduced with permission from Johnson et al. [124].

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2281

DNA structure and function A. Travers and G. Muskhelishvili

ciated with individual base pairs and consequently the

stacking interactions between base pairs (Fig. 4)
[18,19]. In other words, the ability to assume the B-
conformation, which confers on DNA an important
aspect of its unique genetic role, is itself dependent on
base-type and in particular on A–T and G–C base
pairs. Although this might constitute a reason for the
selection of these base pairs in most DNA molecules
no such simple argument can be advanced for the use
of A–U and G–C base pairs in RNA, although even
in RNA the stability of different base-steps and hence
of the double-helix itself is likely dependent on the pre-
cise nature of the constituent base pairs.

Fig. 4. Dipole moments of A–T and G–C base pairs. Reproduced

with permission from Hunter [125].

Alternative modes of sequence

recognition
The gold standard of sequence recognition in the dou-
ble helix employed in both transcription and replica-
tion utilizes the base pairing rules formulated by
Watson and Crick [1]. However, as they acknowl-
edged, bases can pair in different ways. In particular, a
different base-pairing geometry, the Hoogsteen base
pairs, in which the purine base is rotated relative to
that in the standard base pair, reduces the distance
between the C1 carbon atoms of the associated sugar
moieties [20] (Fig. 2B). Although this type of base
pairing is incompatible with the structure of the
canonical DNA double helix, it is found in other
structural forms of DNA, notably H-DNA and G-
quadruplexes [21–24] (see below). Similarly, yet
another type of noncanonical base pair has been
postulated to stabilize the i-motif formed by sequences
complementary to those forming G-quadruplexes [25].
Yet another form of DNA–DNA interaction, which
has received relatively scant attention, is the ability of
two double helices of the same sequence to align with
Fig. 3. Structures of alternative base pairs maintaining Watson–
each other [26,27]. The attractive force causing this
Crick hydrogen bonding. A, adenine; C, cytosine; DAP, DNA self-assembly might function in biological pro-
diaminopurine; G, guanine; H, hypoxanthine; U, uracil. Adapted cesses such as folding of repetitive DNA, recombina-
with permission from Bailly et al. [15]. tion between homologous sequences, and synapsis in

2282 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

meiosis. But how is this type of homologous recogni- narrow minor grooves [31]. Consequently, because in
tion effected? A possible mechanism is the mutual tightly bent DNA, both DNA grooves are narrowed on
alignment of the electrostatic signature of a base the inside of a bend and widened on the outside, G/C-
sequence [28]. Another, not exclusive, suggestion is rich sequences are favoured in which the minor groove
that the flipping out of bases from the double helices points outward and A/T-rich sequences in which the
may also be involved [26]. minor groove points inward.
Even when free in solution, certain DNA sequences
can confer a preferred axial configuration [29,37–39].
DNA as a conformationally flexible
Such sequences contain base-steps that are conforma-
and dynamic polymer
tionally rigid [18,40]; that is, the base-step can adopt
In genomes, DNA molecules are generally very long, only a limited range of conformations. The most nota-
thin polymers with a diameter of 2 nm and a length ble of these sequences are stretches of oligo(dA)(dT) in
that can extend to 108–109 nm. As an information which the AA/TT base-steps are stabilized by bifur-
store, not only must DNA be able to encode the genetic cated hydrogen bonds [41–43]. By themselves, these
information required to specify proteins, but also it sequences are straight, but when juxtaposed with G/C-
should be packaged in a compact form that allows the rich sequences, the whole sequence adopts a curved
accessibility of that information to be regulated. In configuration [44]. Such intrinsically curved molecules
turn, the functional accessing of information may also can, when mimicking the curvature of DNA on the
involve structural changes in the double helix itself. histone octamer, facilitate nucleosome formation [45],
However, the very nature of DNA – again an immen- whereas long straight stretches of oligo(dA)(dT) have
sely long, very thin polymer – requires that within the a lower affinity for the octamer and can serve to phase
cell the molecule be compacted into a small volume nucleosomes in vivo [45].
while maintaining accessibility. These requirements for Not only must DNA be bendable in order to be
compaction, accessibility and structural modulation packaged efficiently, but the copying of the DNA
imply that DNA be both flexible and able to change sequence during transcription, or DNA replication,
conformation in response to enzymatic manipulation. requires the separation of the two strands of the dou-
The DNA molecule may be modelled as an extremely ble helix, a transition in which the double helix is
long thin string of moderate elasticity that can be bent untwisted to form a bubble. The initiation of copying
into the configurations required for packaging. Both the at the points at which strand separation is nucleated is
preferred direction of bending and the stiffness are facilitated by highly localized less stable DNA
sequence dependent [29–31]. A directional bending pref- sequences with lower stacking and melting energies
erence, or bending anisotropy, facilitates the wrapping [46]. Of these, the least thermally stable base-step is
of DNA on a complementary protein surface. However, TpA [47,48]. Such localized untwisting of DNA affects
such a preference also implies that anisotropy increases both DNA melting and also DNA bending. When a
the overall stiffness because it reduces the degrees of bubble is formed in the DNA double helix, not only is
bending freedom. In other words, by reducing bending its bending flexibility increased [49], but so too is the
freedom, the intrinsic bending entropy of the sequence directionality of sequence-directed bending becomes
is reduced and on binding to a preferred protein surface attenuated and more isotropic.
there is a corresponding reduction in the entropic
penalty [32]. Such bending preferences are important
DNA as an energy store
determinants of the binding of both enzymatic manipu-
lators of DNA and the abundant, so-called ‘architec- An often overlooked function of DNA in a cell
tural’, DNA-binding proteins [33], which direct the local nucleus or bacterial nucleoid is that it can act as an
packaging of the polymer. The archetypical example of energy store for facilitating the transit of DNA and
this mode of packaging is the nucleosome core particle – RNA polymerases. This emergent property is a direct
the fundamental unit of DNA packaging in eukaryotic consequence of the double-helical character of the
chromosomes – in which 145 bp of DNA are wrapped molecule. Not only does it exist as a simple intramo-
in 1.6 turns tightly around a histone octamer [34]. The lecular interwound coil, but under torsional stress, the
signature of bending directionality is the presence of DNA chain can adopt a coiled configuration, or super-
alternating short stretches of G/C-rich and A/T-rich coil [50,51]. Such supercoils have a higher intrinsic
DNA sequences in the helical phase [31,35,36]. Such an energy than DNA molecules not subject to torsional
organization confers bending anisotropy because G/C stress. Within both the eukaryotic nucleus and the bac-
and A/T-rich sequences favour, respectively, wide and terial nucleoid supercoiling is ubiquitous [52–54].

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2283

DNA structure and function A. Travers and G. Muskhelishvili

An open circle represents a state in which the DNA lowing one or other of the grooves, rotate along a heli-
molecule, under the prevailing environmental condi- cal path [60]. Many protein complexes are much more
tions, occupies an energetic minimum. It is ‘relaxed’. bulky than DNA and their freedom to rotate around
However, enzymatic manipulation of DNA using the DNA may be constrained by molecular crowding
energy from ATP can alter the torsional state of DNA creating viscous drag [61]. In some cases, the polymeriz-
inducing more coiling – ‘supercoiling’ – within such a ing enzymes may even be restrained in a fixed spatial
closed system. This coiling can be ‘positive’ – over- position by secondary physical attachment to extensive
winding – in the same sense as the DNA double helix structures, such as membranes [62–65]. Under these cir-
or ‘negative’ – underwinding – in the opposite sense cumstances, provided the rotation of the DNA molecule
[51] (Fig. 5). Enzymatic manipulation of DNA super- is also constrained, torsional strain is generated such
coiling can take two forms. On the one hand, enzyme, that the DNA is overwound downstream of the advanc-
termed a ‘topoisomerase’, can bind to a single site and ing enzyme and underwound upstream (Fig. 6). This
directly change the coiling [55] and on the other hand, principle, first proposed by Liu and Wang [56], plays a
coiling can be changed by the movement under partic- key role in the genetic organization of chromosomes.
ular constraints of a protein, or protein complex, such Another important function of topoisomerases is to
as RNA polymerase along the DNA [56–58]. An buffer DNA structure against temperature variations.
example of the first case is DNA gyrase, a bacterial to- Living organisms exist over a broad temperature range
poisomerase that introduces negative supercoiling into of approximately 15 °C to +120 °C and within this
DNA [59] thus facilitating both compaction and strand range, individual organisms can tolerate quite wide
separation at biologically important DNA sequences. temperature variations. At the high end of the overall
In this example, the energy level of DNA is raised. temperature range, the probability of adventitious
Other topoisomerases can reverse this effect, so relax- melting, and consequently of errors in, for example,
ing DNA. transcription initiation, is substantially increased. To
However, although topoisomerases can establish and counteract the possibility of such deleterious bubbles,
maintain an equilibrium state of supercoiling, the pro- extremophiles – for example, thermophilic bacteria
cesses of DNA replication and transcription generate and Archaea – often encode a reverse DNA gyrase
transient changes in DNA supercoiling following the that increases the twist of the DNA double helix.
translocation of the protein complexes along the DNA Indeed, recent calculations suggest such a strategy for
[56]. These transients arise as a direct consequence of stabilizing the double helix would enable the DNA of
the double-helical structure of DNA. When proteins an extremophile, Thermus thermophilus, to retain suffi-
such as RNA polymerase move along DNA, they do cient stability for biological function up to a tempera-
not track linearly along the molecule but instead, by fol- ture of 106 °C, or 15 °C higher, than the maximum
value for free DNA observed in the absence of topo-
logical constraints [66]. Even organisms that exist at
more normal ambient conditions exert fine control
over DNA structure to compensate for temperature
changes. For example, the bacterium Escherichia coli
maintains a constant superhelical stress over a temper-
ature range of 17–37 °C [67]. This effect, presumably
mediated by topoisomerases, compensates for tempera-
ture-dependent alteration of double-helical pitch over
this temperature range.
In both the bacterial nucleoid and the eukaryotic
nucleus, DNA is usually packaged as a negative super-
coil [54] consistent with the preferential binding of
negative supercoils by the most abundant nuclear (the
nucleosome core particle and HMGB proteins) and
Negative Relaxed circle Positive bacterial nucleoid (HU, H-NS and FIS) proteins [33].
supercoil supercoil
But why is DNA packaged in a higher energy state?
Fig. 5. Effect of superhelicity on the conformation of a small DNA Perhaps trivially, as in a ball of string, coiling is an
circle. The figure shows the introduction of positive and negative efficient mode of packaging a long chain and so
cross-overs induced by positive and negative superhelicity, increases the compaction of long stretches of DNA.
respectively. But the storage of negative supercoils also has the

2284 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

Fig. 6. Induction of superhelicity in a

constrained DNA domain by a fixed
elongating DNA translocase moving in the
direction of the arrow. Negative
superhelicity is generated upstream of the
translocating enzyme and positive
superhelicity downstream. (Upper)
Gradients of superhelicity on each side of
the translocase Adapted from Liu and
Wang [56]. (Lower) Different architectural
proteins might interact with the transient
superhelicity. h indicates superhelical
density.

potential to facilitate the passage of DNA and RNA and HMGB proteins constrain negative superhelicity,
polymerases along a DNA template. If DNA is the octamer binds writhed DNA [34] and the HMGB
already packaged as a negative supercoil, release of proteins untwisted DNA [74]. Another protein, the
these negative supercoils might effectively neutralize bacteria nucleoid-associated protein HU constrains
some or all of the positive superhelicity generated by DNA that is likely both writhed and untwisted [75].
an advancing enzyme and so, at least partially counter- This difference is important, as the negative superhelic-
act any inhibitory effect of positive torsion on the pro- ity generated by a DNA translocase is likely highest
cession of the melted bubble within the polymerase closest to the translocase and decays as the distance
complex. After the passage of the polymerase, the neg- from the translocase increases [76]. In this context, it is
ative superhelicity behind the enzyme would facilitate intriguing that protein complexes containing an HMG
the repackaging of DNA [68]. Such a process would domain, such as certain chromatin remodellers [77,78],
conserve negative superhelicity. The positive superhe- and the FACT transcription elongation complex [79–
licity in eukaryotes might be removed by relaxing to- 81] are either translocases themselves, or act in close
poisomerases, such as topoisomerase II, whereas in proximity to a translocation site [82]. Because untwist-
bacteria, DNA gyrase might use ATP to ultimately ing is favoured by higher superhelical densities, it can
increase the average level of negative superhelicity. be speculated that HMG domains stabilize the high
This distinction highlights a fundamental difference nascent negative superhelicity generated by DNA tran-
between bacteria and eukaryotes. By possessing DNA slocases and only subsequently is some of this superhe-
gyrase, whose activity is sensitive to ATP levels [69], licity constrained by the histone octamer.
bacteria – as well as blue–green algae, at least some An additional important facet of translocase func-
mitochondria and chloroplasts [70–72] – can fine-tune tion is the force applied to DNA by the enzyme com-
the negative superhelicity of the genomic DNA so that plexes. For example, a transcribing bacterial RNA
its energy level reflects energy availability from exter- polymerase can exert a force of ~ 20 pN [83], whereas
nal sources. Lacking DNA gyrase in the nucleus, this the maximal force exerted by bacteriophage Φ29 portal
mechanism is not available to eukaryotes. Neverthe- motor is > 100 pN [84]. Forces of this magnitude have
less, they can, and do, by protein binding conserve the the potential to alter the partition of superhelicity
negative superhelicity generated by DNA translocases. between twist and writhe [85]. Notably forces > 3 pN
An increase in unconstrained (not protein-bound) neg- favour twist – both overtwisting and undertwisting –
ative superhelicity on chromatin decompaction [73] rather than writhe [85]. Depending on the distribution
might arise both from the release of constrained super- of the applied force this implies that, for example, bac-
coils or from the selective relaxation of positive supe- terial RNA polymerase could facilitate transcription
rhelicity generated by transcription. by uncoiling a negatively supercoiled DNA plectoneme
An under-appreciated aspect of the constraint of downstream of the transcribing enzyme. Similarly a 4–
DNA superhelicity by DNA-binding proteins is that 5 pN force is required to extend and uncoil the 30 nm
optimal superhelical density for binding is likely pro- chromatin fibre [86]. Again, this is significantly less
tein dependent. Although both the histone octamer than the force exerted by RNA polymerase.

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2285

DNA structure and function A. Travers and G. Muskhelishvili

If packaged DNA constituted an energy store for Alternative DNA structures

modulating gene expression, the nature of the packag-
ing might depend on energy availability. In bacteria, Although a right-handed double helix is the canonical
there is evidence that this is indeed the case. During image of a DNA, it has long been recognized that the
the late stationary phase of growth when the cells are molecule can adopt a number of other biologically
starved of energy, the nucleoid body collapses and the important structures. These include variations on the
DNA is packaged by the highly abundant protein [87]. double helix such as bubbles, Z-DNA, cruciforms and
The proposed mode of packaging by E. coli Dps is slipped loops, three-stranded triple helices (H-DNA)
very different from that of the abundant DNA binding and even distinct four-stranded structures, G-quadru-
proteins characteristic of exponential growth [88] and plexes [21–25,90–93] (Fig. 7). H-DNA demonstrates a
there is no evidence that it constrains DNA superhelic- further structural capability of the DNA. The depar-
ity, suggesting that high superhelical density is not a ture from the standard double-helical structure is
necessary concomitant of compaction. In eukaryotic accompanied by the formation of a different type of
nuclei, as in bacteria, loss of energy results in chroma- base-pair in H-DNA in which the third strand of the
tin compaction [89]. Here, however, the molecular nat- triple stranded complex forms Hoogsteen base pairs
ure of the mechanisms involved is not yet understood. with a Watson–Crick paired double strand in the
Nonetheless, it seems a reasonable proposition that the major groove of the latter [21,22]. Double-stranded
coupling between energy availability and DNA organi- sequences generating a G-quadruplex form a G-rich
zation is tighter in bacteria than in eukaryotes. strand form another structure, the i-motif from the C-

A B C D

E F

Fig. 7. Alternative DNA structures. Gallery of alternative structures showing (A) DNA bubble, (B) Z-DNA, (C) slipped loop, (D) cruciform, (E)
H-DNA, (F) G-quadruplex/i-motif in double-stranded DNA. For the quadruplex/i-motif the structure assumed by the i-motif is likely pH
dependent [19].

2286 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

rich strand [25]. In the quadruplex, the four strands Negative superhelicity also facilitates the formation
are connected by Hoogsteen base pairing [23,24], of other alternative DNA structures, notably the left-
whereas the i-motif has yet another type of noncanoni- handed Z-DNA and also cruciforms, as well as their
cal pairing [25]. slipped loop variants [91,92,99,100]. Although there
In addition to structures that can form over the nor- has been debate about the occurrence of such struc-
mal range of physiological conditions, yet more differ- tures in vivo, in some examples, sequences with the
ent structures can be induced by the application of potential to form these structures are located in the
extensive and torsional forces greater than those nor- vicinity of promoter regions [100,101]. G-Quadruplexes
mally encountered in the cell [94]. Such structures are constitute another class of structure. They form from
very different to the classical B-type double helix [95– a G-rich single strand with a particular sequence orga-
97] and include the underwound S-DNA with an esti- nization and constitute the major structural motif at
mated 33 bp per turn and the overwound P-DNA with the single-stranded termini of eukaryotic telomeres
2.7 bp per turn [94,97], the latter possibly correspond- [20,21]. Additionally, such sequences are frequently
ing to the ‘inside-out’ structure originally proposed by found in double-stranded form in internal positions,
Pauling and Corey [98]. The formation of S-DNA again often located close to promoter regions [102].
requires an extensive force of at least 60 pN, whereas Again, like Z-DNA and slipped loops, their formation
P-DNA is only stable at a positive torque in excess of and that of the complementary C-rich hairpin is pro-
30 pNnm1 [95]. moted by negative superhelicity [103–105].
The formation of alternative structures under physi- The multiplicity of alternative DNA structures whose
ological conditions is favoured not only by particular formation is dependent on the intrinsic torsional stress
sequence organizations and base compositions, but in the DNA begs the question of their function, if any.
also by the energetic environment of these DNA Their frequent association with promoter regions
sequences. One of the simplest, and among the biologi- implies that they might facilitate, but not necessarily be
cally most important, of these alternative structures is essential for, transcription initiation. Although this
a DNA bubble in which strand separation occurs over process is presented as relatively simple in textbooks, in
a short sequence of base pairs generating a region con- reality, the progression from polymerase binding to the
sisting of two separated strands bounded by more sta- escape of an actively transcribing polymerase presents
ble double-helical stretches. Bubble formation is conflicting topological problems (Fig. 8). After bind-
strongly sequence dependent, occurring most fre- ing, the enzyme first mediates the melting of approxi-
quently at the TpA base-step [46] – the least stable of mately slightly more than one turn of double-stranded
all 10 steps [47,48] (Table 1). Consequently, melting is DNA, thereby constraining negative superhelicity. But
favoured by agents – higher temperatures and negative in a closed system this melting of one double-helical
superhelicity – that promote the unwinding of the dou- turn must be balanced by the generation of an equal
ble helix, and the distribution of preferred melting sites and opposite positive superhelicity, which will be most
closely correlates with genetic function. For example, manifest in the immediate vicinity of the polymerase.
DNA sequences in close proximity to the transcription In the absence of abundant topoisomerases relaxing
start point are, on average, enriched in the TpA base- this positive superhelicity, it might be absorbed by
step, as also are specific recombination sites [46]. alternative DNA structures converting them back to a
simple double helix. However, there is another poten-
tial barrier to initiation – the escape of the polymerase.
Table 1. Melting energies of the 10 base-steps. Reproduced from
If the DNA is initially relaxed as the polymerase begins
Protozanova et al. [48].
to move away from the promoter, the advancing com-
Base step Melting energy (kcalmol1) plex will generate negative superhelicity behind and
positive superhelicity in front. Absorption of the nega-
TA 0.12
TG/CA 0.78 tive superhelicity by a DNA sequence with the poten-
AA/TT 1.04 tial to form an alternative structure might then enable
AT 1.27 the release of the polymerase complex from the pro-
AG/CT 1.29 moter. One possible function of these structures is thus
CG 1.44 to act as a torsional buffer [104–108]. Similarly, in
GA/TC 1.66
eukaryotic chromosomes, the positive superhelicity
GG/CC 1.97
in front of polymerase might contribute to the
AC/GT 2.04
GC 2.70 facilitation of the unwrapping of DNA around
nucleosomes [68].

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2287

DNA structure and function A. Travers and G. Muskhelishvili

Fig. 8. A topological machine: topology of

RNA polymerase initiation and escape. The
enzyme (red ellipse) binds at a promoter
site, melts the DNA over a limited region
and then escapes from the promoter.
These transitions are facilitated by a
transcription factor binding upstream of
and contacting the enzyme and are
accompanied by changes in the path of
the DNA around the polymerase and in
the region upstream of the enzyme.
Adapted from Muskhelishvili and Travers
[108].

Although the ability of DNA to assume a variety of yeast origins of DNA replication. In at least the for-
alternative structures is a highly visible manifestation mer case, superhelicity favours strand separation at
of conformational flexibility, the molecule can undergo these sites [46]. By contrast, regions of low melting
other more subtle biologically important transitions. energy DNA extending to ~ 100 bp or more exist both
One such is the property of coiling or writhing under upstream of strongly transcribed E. coli genes
torsional stress. Such coiling is favoured by increasing [110,111] and downstream of many budding yeast
flexibility and hence higher A/T contents and, impor- genes. In both cases, superhelicity introduced by DNA
tantly, can be induced by both negative and positive gyrase and/or by transcription is believed to induce
superhelicity. Negative superhelicity also induces local- DNA writhing [109]. The distinction between writhing
ized strand separation [46] and, in this case, any choice and strand separation is crucial for biological function.
between writhing and strand separation will likely Supercoiled DNA in E. coli has an average superheli-
depend on the both the length and the organization of cal density of 0.05; that is, an average change of
the sequence. In general, for a DNA stretch of low linking number from the fully relaxed form of 1/
melting energy, the longer the sequence the greater the 200 bp. Typically in the very active stable RNA pro-
probability of responding to negative superhelicity by moters of E. coli the 10 hexamer is flanked by blocks
writhing, rather than by melting. This is because DNA of G/C-rich sequence of higher average melting tem-
superhelicity is not distributed uniformly along a perature [109]. Such blocks potentially act as barriers
DNA molecule but instead will be, on average, and so can serve to localize the effects of torque –
localized to those sequences that are most deformable. dependent on the intrinsic superhelicity or on direct
If a highly deformable sequence is short, and particu- enzymatic manipulation – to the short 10 region.
larly if it is flanked by sequences for which deforma- The consequence is that the local superhelical density
tion is energetically unfavourable, available within this sequence is much higher than the average
superhelicity will be effectively concentrated in the superhelical density and so melting is favoured [46].
short sequence resulting in a high local superhelical This contrasts with the upstream regions of such pro-
density [7,8,109]. By contrast, for a longer deformable moters, which although also sensitive to supercoiling
sequence, the available superhelicity will be spread [111], are much more extensive and more uniform in
more extensively. Exemplars of short deformable base composition. Consequently, any superhelicity is
sequences are the hexameric 10 region of strong bac- distributed over more double-helical turns resulting in
terial promoters and similar sequences within budding a lower local superhelical density than might be experi-

2288 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

enced in a ‘gated’ 10 region, and thus for a prefer- rich and most thermodynamically unstable DNA
ence for writhing rather than melting. The localization sequences in the Saccharomyces cerevisiae genome are
of linking number changes in supercoiled DNA is also located in 30 flanking regions [110]. This distribution
an essential element in the formation of such struc- of base composition on a genomic scale implies that,
tures as cruciforms and slipped loops [91,92]. on average, coding sequences are stiffer or less bend-
The sequence-dependent bending anisotropy and able, whereas noncoding sequences are both more
elasticity of DNA have the potential to organize a flexible and more susceptible to strand separation.
DNA molecule in a preferred writhing configuration, However, in apparent contradiction to these varia-
or configurations, under superhelical stress. Such pre- tions in flexibility, in eukaryotic chromosomes coding
ferred configurations might contribute to the distinc- sequences have a higher nucleosome occupancy than
tion of discrete domains within a single DNA noncoding sequences [45,112]. But, again, this pattern
molecule (e.g. (Fig. 9). When stabilized by proteins, of occupancy is possibly related to another sequence-
such structures, which in an unconstrained DNA mol- dependent physical property of the polymer, the
ecule are likely dynamic, might act as separate topo- higher intrinsic entropy of certain A/T-rich sequences
logical domains and enable differential modes of gene [32].
regulation within each domain. We suggest that effects The occurrence of the more A/T-rich sequences in
such as these would be the expression of analogue the flanking regions of genes has functional signifi-
information spanning several kilobases. cance. At the 50 -end of a transcription unit there is an
obvious correlation with the requirement for RNA
polymerase to melt DNA prior to transcription initia-
DNA and genetic organization
tion. But at the 30 -ends of transcription units, polymer-
The physicochemical properties conferred by DNA ase dissociates and releases the constrained unwound
sequence not only determine bending and melting DNA so that it reforms a double helix. One possibility
preferences, but also correlate strongly with the is that such regions serve as topological sinks, absorb-
genetic organization of both eukaryotic and bacterial ing by writhing any positive superhelicity generated in
chromosomes. In general, the coding sequences of advance of the transcribing enzyme. This would block
genes have a G/C-rich bias [45,112]. In part, this is the transmission of any such superhelicity to a neigh-
because the codons for the most abundant amino bouring gene with the potential for disrupting its chro-
acids also have a G/C-rich bias [113,114]. The corol- matin structure. Instead, the writhed DNA would
lary is that noncoding DNA sequences, including in- serve as an appropriate substrate for relaxation by to-
trons as well as 50 and 30 flanking DNA sequences, poisomerases; in particular, topoisomerase II, which is
are generally more A/T-rich. Indeed the most A/T- preferentially associated with actively transcribed genes
[115]. Topoisomerase II, together with topoisomer-
ase I, is also found in regions of low nucleosome occu-
pancy at promoters [115]. However, measurement of
the association of topoisomerase II with its optimal
binding sites is precluded because of their highly repet-
itive and redundant nature.
The relationship of the physicochemical properties
of DNA to chromosome organization and function in
not only apparent at the level of individual genes and
transcription units, but is also a feature of whole
bacterial chromosomes. These chromosomes comprise,
in general, a single circular DNA molecule which can
vary in length from ~ 0.5 Mb to 6–10 Mb. Remark-
ably in these chromosomes, at least in most c-Proteo-
bacteria, gene order is highly conserved such that
those genes that are highly expressed during exponen-
Fig. 9. Transient configuration of a negatively supercoiled mini
tial growth are clustered near the origin of DNA repli-
chromosome. AFM visualization of a circular supercoiled pBR322
DNA molecule. In this configuration the molecule exhibits a distinct
cation, whereas those that are more active during
domain structure delimited by the duplex–duplex contacts in the episodes of environmental stress resulting in the cessa-
centre of the picture. Previously unpublished image reproduced tion of growth are more frequent in the vicinity of the
with kind permission from Sebastian Maurer. replication termini [116,117]. However, not only is

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2289

DNA structure and function A. Travers and G. Muskhelishvili

there a gradient of gene organization from origin to matic manipulation, how are they integrated with the
terminus, but also this gradient correlates, on average, primary function of DNA as an information store?
with a gradient of base composition so that in each Like DNA, a fundamental property of RNA is to
replichore the most stable, G/C-rich, DNA is close to encode protein sequences, but unlike a DNA genome,
the origin while the least stable is at the terminus an RNA genome must combine both information
[117]. This average pattern of course includes wide storage and its functional expression during the trans-
variations at the level of individual genes. Yet another lation of the nucleotide sequence into protein. These
feature that exhibits a graded response from origin to two requirements are not necessarily wholly compati-
terminus is the distribution of binding sites for DNA ble. For example, when there are strong selective
gyrase [116,118], a topoisomerase that inserts negative pressures to maintain the integrity of the genomic
superhelical turns into DNA [55]. Again these are con- nucleotide sequence, the option of regulating transla-
centrated primarily in proximity to the origin of repli- tion by modulating the half-life of an RNA molecule
cation and thus create the potential for the DNA in is effectively excluded. Perhaps more tellingly, in
this region to be more highly negatively supercoiled known present-day biological systems, RNA mole-
than that close to the terminus. This overall pattern of cules that function as messengers, and also as ge-
organization can couple chromosome structure to nomes (e.g. those of poliovirus and Qb
energy availability [69]. When bacteria are shifted to a bacteriophage) are, relative to most genomic DNA
fresh rich growth medium, ATP levels rise, activating molecules, very short, comprising only a few thou-
DNA gyrase and thus increasing the negative superhe- sand nucleotides. An advantage of the separation of
lical density of the chromosome [60]. This would be responsibilities between the two types of polynucleo-
localized to the origin-proximal region and would, in tide is that the juxtaposition and catenation of indi-
turn, activate the genes producing the necessary com- vidual genes into much longer molecules enhances the
ponents for growth – the transcription and translation potential regulatory repertoire of gene expression. In
machinery – as well as providing an appropriate envi- particular, the coordination of gene expression can be
ronment for DNA replication. Once DNA replication facilitated at the local level by the structural interplay
is initiated, the passage of the replisomes along the arising from the transcriptional activity of adjacent
two replichores would by itself generate a gradient of genes, depending on whether they are organized in
superhelicity by the Liu/Wang principle [56], with the tandem or transcription is convergent or divergent.
more negatively supercoiled DNA again being located At a higher level of structural organization the conti-
closer to the origin and the more relaxed DNA close nuity afforded by a single DNA double helix in a
to the terminus. Again, by analogy to transcriptions, chromosome permits the organization of genes, and
the DNA close to the terminus, in concert with topoi- hence the available DNA information, into distinct
somerases, might act as a topological barrier between structural and functional domains comprising many
the two replichores. The bacterial chromosome thus protein-coding elements. Apart from these consider-
functions as a topological machine in which the overall ations, the ability of DNA to accrete more and more
distribution of DNA sequences reflects the coupling packets of information – as genes or as regulatory
between the processing of the replisomes and gene elements – into longer and longer chromosomal mole-
expression. cules, is arguably an important factor contributing to
Although the Liu/Wang principle was initially increases in organismal complexity. Indeed, the postu-
conceived as applying to naked DNA, it is equally lated usurping of RNA by DNA as the informational
valid when considered in the context of higher order store within a cell by itself epitomizes an evolutionary
structures generated by DNA packaging. In eukaryotic increase in biological complexity; one that enables
nuclei, despite the existence of the 30 nm fibre in vivo increased possibilities for information storage – such
being recently questioned [118], the left-handed coiling as the differential encoding on complementary strands
of the nucleosome stacks responds to torsional forces – – and processing and hence, provides a substrate for
such as those generated by transcription – by unwind- further natural selection.
ing on application of positive torsion and correspond- Any increase in the length of chromosomal DNA
ingly rewinding with applied negative torsion [119]. molecules should be coupled to – and possibly limited
by – mechanisms for the generation and maintenance of
genomic integrity. In this context, a relevant biological
Informational capacity
example is provided by ciliates – a group of single-celled
Although the physicochemical properties of DNA organisms including the causal agents of malaria and
determine its dynamic roles in the context of enzy- sleeping sickness – in which regulation of DNA-directed

2290 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

gene expression is separated from maintenance of the 5 von Neumann J (1958) The Computer and the Brain.
germ line. These organisms contain two types of nuclei. Yale University Press, New Haven, CT.
One, the micronucleus, containing the complete diploid 6 Marr C, Geertz M, H€ utt MT & Muskhelishvili G
genome, serves as the germ line and does not express (2008) Dissecting the logical types of network control
genes [120], whereas the second, the polyploid macronu- in gene expression profiles. BMC Syst Biol 2, 18.
cleus contains a highly edited and fragmented version of 7 Harris SA, Laughton CA & Liverpool TB (2008)
the genome, and serves as the vehicle controlling gene Mapping the phase diagram of the writhe of DNA
expression [121,122]. Only the micronucleus undergoes nanocircles using atomistic molecular dynamics
simulations. Nucleic Acids Res 36, 21–29.
mitotic chromosomal segregation [123].
8 Sayar M, Avsßaroglu B & Kabakcßioglu A (2010) Twist-
writhe partitioning in a coarse-grained DNA minicircle
Conclusions model. Phys Rev E Stat Nonlin Soft Matter Phys 81,
041916.
Although, quite rightly, the ability of DNA to code for
9 Arnott S, et al. (1966) Molecular and crystal structures
protein sequences is often emphasized, equally impor- of double- helical RNA. III. An 11-fold molecular
tant is the encoding of information that enables both model and comparison of the agreement between the
the packaging of the polymer and the regulation of the observed and calculated three- dimensional diffraction
expression and accessibility of the protein-coding infor- data for 10 and 11-fold models. J Mol Biol 27,
mation. This latter property depends to a much greater 535–548.
extent on the sequence-dependent physicochemical 10 Arnott S, Hutchinson F, Spencer M, Wilkins MH,
properties of the molecule and thus on the cumulative Fuller W & Langridge R (1966) X-ray diffraction
properties of a succession of a small number of base studies of double helical ribonucleic acid. Nature 211,
pairs. This analogue characteristic contrasts with the 227–232.
digital encoding of protein sequences. A further crucial 11 Dickerson RE, Drew HR, Conner BN, Wing RM,
aspect of DNA function is the dynamic nature of the Fratini AV & Kopka ML (1982) The anatomy of A-,
structural transitions observed during its replication B-, and Z-DNA. Science 216, 475–485.
and transcription. These transitions involve both devia- 12 Arnott S (2006) Historical article: DNA polymorphism
tions from the canonical double helix and also topologi- and the early history of the double helix. Trends
cal transformations of the trajectory of the double helix Biochem Sci 31, 349–354.
facilitating both packaging and regulation. Taken 13 Franklin RE & Gosling RG (1953) Molecular
together, these properties enable DNA to function as a configuration in sodium thymonucleate. Nature 171,
supremely efficient and versatile coding device. 740–741.
14 Horme~ no S, Ibarra B, Carrascosa JL, Valpuesta JM,
Moreno-Herrero F & Arias-Gonzalez JR (2011)
Author contributions Mechanical properties of high-G.C content DNA with
A-type base-stacking. Biophys J 100, 1996–2005.
Both authors contributed sections to the manuscript.
15 Bailly C, Waring MJ & Travers AA (1995) Effects of
base substitutions on the binding of a DNA-bending
References protein. J Mol Biol 253, 1–7.
16 Virstedt J, Berge T, Henderson RM, Waring MJ &
1 Watson JD & Crick FH (1953) Molecular structure of Travers AA (2004) The exocyclic purine 2–amino
nucleic acids; a structure for deoxyribose nucleic acid. group influences the stiffness of the DNA double helix.
Nature 171, 737–738. J Struct Biol 148, 66–85.
2 Avery OT, Macleod CM & McCarty M (1944) Studies 17 Tseng YD, Ge H, Wang X, Edwardson JM, Waring
on the chemical nature of the substance inducing MJ, Fitzgerald WJ & Henderson RM (2005) Atomic
transformation of pneumococcal types: induction of force microscopy study of the structural effects
transformation by a desoxyribonucleic acid fraction induced by echinomycin binding to DNA. J Mol Biol
isolated from pneumococcus type III. J Exp Med 79, 345, 745–758.
137–158. 18 Hunter CA (1993) Sequence-dependent DNA
3 Chargaff E, Lipshitz R, Green C & Hodes ME (1951) structure. The role of base stacking interactions. J Mol
The composition of the deoxyribonucleic acid of Biol 230, 1025–1054.
salmon sperm. J Biol Chem 192, 223–230. 19 Gallego J, Luque FJ, Orozco M, Burgos C, Alvarez-
4 deChadarevian S & Kamminga H (2002) Builla J, Rodrigo MM & Gago F (1994) DNA
Representations of the Double Helix. Whipple sequence-specific reading by echinomycin: role of
Museum of the History of Science, Cambridge, UK.

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2291

DNA structure and function A. Travers and G. Muskhelishvili

hydrogen bonding and stacking interactions. J Med 35 Travers AA & Klug A (1987) The bending of DNA in
Chem 7, 1602–1609. nucleosomes and its wider implications. Philos Trans R
20 Hoogsteen K (1963) The crystal and molecular Soc Lond B 317, 537–561.
structure of a hydrogen-bonded complex between 1– 36 Gartenberg MR & Crothers DM (1988) DNA
methylthymine and 9–methyladenine. Acta Crystallogr sequence determinants of CAP-induced bending and
16, 907–916. protein binding affinity. Nature 333, 824–829.
21 Htun H & Dahlberg JE (1988) Single strands, triple 37 Ntambi JM, Marini JC, Bangs JD, Hajduk SL,
strands, and kinks in H-DNA. Science 241, 1791– Jimenez HE, Kitchin PA, Klein VA, Ryan KA &
1796. Englund PT (1984) Presence of a bent helix in
22 Harvey SC, Luo J & Lavery R (1988) DNA stem-loop fragments of kinetoplast DNA minicircles from several
structures in oligopurine-oligopyrimidine triplexes. trypanosomatid species. Mol Biochem Parasitol 12,
Nucleic Acids Res 16, 11795–11809. 273–286.
23 Sundquist WI & Klug A (1989) Telomeric DNA 38 Diekmann S (1986) Sequence specificity of curved
dimerizes by formation of guanine tetrads between DNA. FEBS Lett 195, 53–56.
hairpin loops. Nature 342, 825–829. 39 Brukner I, Diakic M, Savic A, Susic S, Pongor S &
24 Kang C, Zhang X, Ratliff R, Moyzis R & Rich A Suck D (1993) Evidence for opposite groove-directed
(1992) Crystal structure of four-stranded Oxytricha curvature of GGGCCC and AAAAAA tracts. Nucleic
telomeric DNA. Nature 356, 126–131. Acids Res 21, 1025–1029.
25 Lyamichev VI, Mirkin SM & Frank-Kamenetskii MD 40 Olson WK, Gorin AA, Lu XJ, Hock LM & Zhurkin
(1985) A pH-dependent structural transition in the VB (1998) DNA sequence-dependent deformability
homopurine-homopyrimidine tract in superhelical deduced from protein–DNA crystal complexes. Proc
DNA. J Biomol Struct Dyn 3, 327–338. Natl Acad Sci USA 95, 11163–11168.
26 Inoue S, Sugiyama S, Travers AA & Ohyama T (2007) 41 Coll M, Frederick CA, Wang AH & Rich A (1987) A
Self-assembly of double-stranded DNA molecules at bifurcated hydrogen-bonded conformation in the d
nanomolar concentrations. Biochemistry 46, (A.T) base pairs of the DNA dodecamer d
164–171. (CGCAAATTTGCG) and its complex with
27 Baldwin GS, Brooks NJ, Robson RE, Wynveen A, distamycin. Proc Natl Acad Sci USA 84, 8385–8389.
Goldar A, Leikin S, Seddon JM & Kornyshev AA 42 Nelson HC, Finch JT, Luisi BF & Klug A (1987) The
(2008) DNA double helices recognize mutual sequence structure of an oligo(dA).oligo(dT) tract and its
homology in a protein free environment. J Phys Chem biological implications. Nature 330, 221–226.
B 112, 1060–1064. 43 DiGabriele AD, Sanderson MR & Steitz TA (1989) A
28 Kornyshev AA & Leikin S (2001) Sequence DNA dodecamer containing an adenine tract
recognition in the pairing of DNA duplexes. Phys Rev crystallizes in a unique lattice and exhibits a new bend.
Lett 86, 3666–3669. Proc Natl Acad Sci USA 86, 1816–1820.
29 Koo HS, Wu HM & Crothers DM (1986) DNA bending 44 Ulanovsky L, Bodner M, Trifonov EN & Choder M
at adenine–thymine tracts. Nature 320, 501–506. (1986) Curved DNA: design, synthesis, and
30 Zhurkin VB (1983) Specific alignment of nucleosomes circularization. Proc Natl Acad Sci USA 83, 862–866.
on DNA correlates with periodic distribution of 45 Struhl K & Segal E (2013) Determinants of nucleosome
purine-pyrimidine and pyrimidine-purine dimers. FEBS positioning. Nat Struct Mol Biol 20, 267–273.
Lett 158, 293–297. 46 Drew HR, Weeks JR & Travers AA (1985) Negative
31 Satchwell SC, Drew HR & Travers AA (1986) supercoiling induces spontaneous unwinding of a
Sequence periodicities in chicken nucleosome core bacterial promoter. EMBO J 4, 1025–1032.
DNA. J Mol Biol 191, 659–675. 47 SantaLucia J Jr (1998) A unified view of polymer,
32 Travers AA, Muskhelishvili G & Thompson JMT dumbbell and oligonucleotide nearest-neighbor
(2012) DNA information; from digital code to thermodynamics. Proc Natl Acad Sci USA 95, 1460–
analogue structure. Philos Trans R Soc Lond A 370, 1465.
2960–2986. 48 Protozanova E, Yakovchuk P & Frank-Kamenetskii
33 Luijsterburg MS, White MF, van Driel R & Dame RT MD (2004) Stacked–unstacked equilibrium at the nick
(2008) The major architects of chromatin: architectural site of DNA. J Mol Biol 342, 775–785.
proteins in bacteria, archaea and eukaryotes. Crit Rev 49 Yan J & Marko J (2004) Localized single-stranded
Biochem Mol Biol 43, 393–418. bubble mechanism for cyclization of short double helix
34 Luger K, M€ader AW, Richmond RK, Sargent DF & DNA. Phys Rev Lett 93, 108108.
Richmond TJ (1997) Crystal structure of the 50 Vinograd J, Lebowitz J, Radloff R, Watson R &
nucleosome core particle at 2.8 A resolution. Nature Laipis P (1965) The twisted circular form of polyoma
389, 251–260. viral DNA. Proc Natl Acad Sci USA 53, 1104–1111.

2292 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

51 Cozzarelli NR, Boles TC & White JH (1990) Primer coli during coupled transcription-translation of genes
on the topology and geometry of DNA supercoiling. encoding integral inner membrane polypeptides. J Biol
In DNA Topology and its Biological Effects Chem 269, 15362–15370.
(Cozzarelli NR & Wang JC, eds), pp. 139–184. Cold 66 Meyer S, Jost D, Theodorakopoulos N, Peyrard M,
Spring Harbor Laboratory Press, NY. Lavery R & Everaers R (2013) Temperature
52 Stonington OG & Pettijohn DE (1971) The folded dependence of the DNA double helix at the nanoscale:
genome of Escherichia coli isolated in a protein–DNA- structure, elasticity, and fluctuations. Biophys J 105,
RNA complex. Proc Natl Acad Sci USA 68, 6–9. 1904–1914.
53 Worcel A & Burgi E (1972) On the structure of the 67 Goldstein E & Drlica K (1984) Regulation of bacterial
folded chromosome of Escherichia coli. J Mol Biol 71, DNA supercoiling: plasmid linking numbers vary with
127–147. growth temperature. Proc Natl Acad Sci USA 81,
54 Bauer W (1978) Structure and reactions of closed 4046–4050.
duplex DNA. Ann Rev Biophys Bioeng 7, 287–313. 68 Studitsky VM, Clark DJ & Felsenfeld G (1994) A
55 Wang JC (2002) Cellular roles of DNA histone octamer can step around a transcribing
topoisomerases: a molecular perspective. Nat Rev Mol polymerase without leaving the template. Cell 76, 371–
Cell Biol 3, 430–440. 382.
56 Liu LF & Wang JC (1987) Supercoiling of the DNA 69 van Workum M, van Dooren SJ, Oldenburg N,
template during transcription. Proc Natl Acad Sci Molenaar D, Jensen PR, Snoep JL & Westerhoff HV
USA 84, 7024–7027. (1996) DNA supercoiling depends on the
57 Wu HY, Shyy S, Wang JC & Liu LF (1988) phosphorylation potential in Escherichia coli. Mol
Transcription generates positively and negatively Microbiol 20, 351–360.
supercoiled domains in the template. Cell 53, 433–440. 70 Castora FJ, Vissering FF & Simpson M (1983) The
58 Tsao YP, Wu HY & Liu LF (1989) Transcription- effect of bacterial DNA gyrase inhibitors on DNA
driven supercoiling of DNA: direct biochemical synthesis in mammalian mitochondria. Biochim
evidence from in vitro studies. Cell 56, 111–118. Biophys Acta 740, 417–427.
59 Gellert M, Mizuuchi K, O’Dea MH & Nash HA 71 Itoh R, Takahashi H, Toda K, Kuroiwa H & Kuroiwa
(1976) DNA work: an enzyme that introduces T (1997) DNA gyrase involvement in chloroplast-
superhelical turns into DNA. Proc Natl Acad Sci USA nucleoid division in Cyanidioschyzon merolae. Eur J
3, 3872–3876. Cell Biol 73, 252–258.
60 Harada Y, Ohara O, Takatsuki A, Itoh H, Shimamoto 72 Wall MK, Mitchenall LA & Maxwell A (2004)
N & Kinosita K Jr (2001) Direct observation of DNA Arabidopsis thaliana DNA gyrase is targeted to
rotation during transcription by Escherichia coli RNA chloroplasts and mitochondria. Proc Natl Acad Sci
polymerase. Nature 409, 113–115. USA 101, 7821–7826.
61 Wang JC & Liu LF (1990) DNA replication: 73 Naughton C, Avlonitis N, Corless S, Prendergast
topological aspects and the roles of DNA JG, Mati IK, Eijk PP, Cockroft SL, Bradley M,
topoisomerases. In DNA Topology and its Biological Ylstra B & Gilbert N (2013) Transcription forms
Effects (Cozzarelli NR & Wang JC, eds), pp. 321–340. and remodels supercoiling domains unfolding large-
Cold Spring Harbor Laboratory Press, New York. scale chromatin structures. Nat Struct Mol Biol 20,
62 Lodge J, Kazik T & Berg DE (1989) Formation of 387–395.
supercoiling domains in plasmid pBR322. J Bacteriol 74 Murphy FV 4th, Sweet RM & Churchill MEA
171, 2181–2187. (1999) The role of intercalating residues in
63 Lynch AS & Wang JC (1993) Anchoring of DNA to chromosomal high-mobility-group protein DNA
the bacterial cytoplasmic membrane through binding, bending and specificity. EMBO J 18,
cotranscriptional synthesis of polypeptides encoding 6610–6618.
membrane proteins or proteins for export: a 75 Guo F & Adhya S (2007) Spiral structure of
mechanism of plasmid hypernegative supercoiling in Escherichia coli HUab provides foundation for
mutants deficient in DNA topoisomerase I. J Bacteriol DNA supercoiling. Proc Natl Acad Sci USA 104,
175, 1645–1655. 4309–4314.
64 Bowater RP, Chen D & Lilley DMJ (1994) Elevated 76 Meyer S & Beslon G (2014) Torsion-mediated
unconstrained supercoiling of plasmid DNA generated interaction between adjacent genes. PLoS Comput Biol
by transcription and translation of the tetracycline 10, e1003785.
resistance gene in eubacteria. Biochemistry 33, 9266– 77 Mohrmann L & Verrijzer CP (2005) Composition and
9275. functional specificity of SWI2/SNF2 class chromatin
65 Ma D, Cook DN, Pon NG & Hearst JE (1994) remodeling complexes. Biochim Biophys Acta 1681, 59–
Efficient anchoring of RNA polymerase in Escherichia 73.

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2293

DNA structure and function A. Travers and G. Muskhelishvili

78 Bartholomew B (2014) Regulating the chromatin 92 Mace HAF, Pelham HRB & Travers AA (1983)
landscape: structural and mechanistic perspectives. Association of an S1 nuclease-sensitive structure with
Annu Rev Biochem 83, 671–696. short direct repeats 50 of Drosophila heat shock genes.
79 Orphanides G, Wu WH, Lane WS, Hampsey M & Nature 304, 555–557.
Reinberg D (1999) The chromatin-specific 93 Minyat EE, Khomyakova EB, Petrova MV, Zdobnov
transcription elongation factor FACT comprises EM & Ivanov VI (1995) Experimental evidence for
human Spt16 and SSRP1 proteins. Nature 400, slipped loop DNA, a novel folding type for
284–288. polynucleotide chain. J Biomol Struct Dyn 13,
80 Brewster NK, Johnston GC & Singer RA (2001) A 523–527.
bipartite yeast SSRP1 analog comprised of Pob3 and 94 Bustamante C, Bryant Z & Smith SE (2003) Ten years
Nhp6 proteins modulates transcription. Mol Cell Biol of tension: single molecule DNA mechanics. Nature
21, 3491–3502. 421, 423–427.
81 Formosa T, Eriksson P, Wittmeyer J, Ginn J, Yu Y & 95 Smith SB, Finzi L & Bustamante C (1992) Direct
Stillman DJ (2001) Spt16-Pob3 and the HMG protein measurements of the elasticity of single DNA
Nhp6 combine to form the nucleosome-binding factor molecules using elastic beads. Science 258, 1122–1126.
SPN. EMBO J 2, 3506–3517. 96 Strick TR, Allemand JF, Bensimon D, Lavery R &
82 Mason PB & Struhl K (2003) The FACT complex Croquette V (1996) The elasticity of a single
travels with elongating RNA polymerase II and is supercoiled DNA molecule. Science 271, 1835–1837.
important for the fidelity of transcriptional initiation 97 Allemand JF, Bensimon D, Lavery R & Croquette V
in vivo. Mol Cell Biol 23, 8323–8333. (1998) Stretched and overwound DNA forms a
83 Yin H, Wang MD, Svoboda K, Landick R, Block SM Pauling-like structure with exposed bases. Proc Natl
& Gelles J (1995) Transcription against an applied Acad Sci USA 95, 14152–14157.
force. Science 270, 1653–1657. 98 Pauling L & Corey RB (1953) A proposed structure
84 Rickgauer JP, Fuller DN, Grimes S, Jardine PJ, for nucleic acids. Proc Natl Acad Sci USA 39, 84–97.
Anderson DL & Smith DE (2008) Portal motor velocity 99 Singleton CK, Klysik J, Stirdivant SM & Wells RD
and internal force resisting viral DNA packaging in (1982) Left-handed Z-DNA is induced by supercoiling
bacteriophage U29. Biophys J 94, 159–167. in physiological ionic conditions. Nature 299, 312–316.
85 Strick TR, Allemand JF, Bensimon D & Croquette V 100 Glaser RL, Thomas GH, Siegfried E, Elgin SC & Lis
(1998) Behavior of supercoiled DNA. Biophys J 74, JT (1990) Optimal heat-induced expression of the
2016–2028. Drosophila hsp26 gene requires a promoter sequence
86 Kruithof M, Chien FT, Routh A, Logie C, Rhodes D containing (CT)n. (GA)n repeats. J Mol Biol 211,
& van Noort J (2009) Single-molecule force 751–761.
spectroscopy reveals a highly compliant helical folding 101 Schroth GP, Chou PJ & Ho PS (1992) Mapping Z-
for the 30-nm chromatin fiber. Nat Struct Mol Biol 16, DNA in the human genome. Computer-aided mapping
534–540. reveals a nonrandom distribution of potential Z–
87 Wolf SG, Frenkiel D, Arad T, Finkel SE, Kolter R & DNA-forming sequences in human genes. J Biol Chem
Minsky A (1999) DNA protection by stress-induced 267, 11846–11855.
biocrystallization. Nature 400, 83–85. 102 Huppert JL & Balasubramanian S (2007) G-
88 Grant RA, Filman DJ, Finkel SE, Kolter R & Hogle quadruplexes in promoters throughout the human
JM (1998) The crystal structure of Dps, a ferritin genome. Nucleic Acids Res 35, 406–413.
homolog that binds and protects DNA. Nat Struct 103 Voloshin ON, Veselkov AG, Belotserkovskii BP,
Biol 5, 294–303. Danilevskaya ON, Pavlova MN, Dobrynin VN &
89 Visvanathan A, Ahmed K, Even-Faitelson L, Lleres Frank-Kamenetskii MD (1992) An eclectic DNA
D, Bazett-Jones DP & Lamond AI (2013) Modulation structure adopted by human telomeric sequence under
of higher order chromatin conformation in superhelical stress and low pH. J Biomol Struct Dyn 9,
mammalian cell nuclei can be mediated by polyamines 643–652.
and divalent cations. PLoS One 8, e67689. 104 Kouzine F, Sanford S, Elisha-Feil Z & Levens D
90 Crawford JL, Kolpak FJ, Wang AH, Quigley GJ, (2008) The functional response of upstream DNA to
van Boom JH, van der Marel G & Rich A (1980) dynamic supercoiling. Nat Struct Mol Biol 15,
The tetramer d(CpGpCpG) crystallizes as a left- 146–154.
handed double helix. Proc Natl Acad Sci USA 77, 105 Sun D & Hurley LH (2009) The importance of
4016–4020. negative superhelicity in inducing the formation of G-
91 Lilley DMJ (1980) The inverted repeat as a quadruplex and i-motif structures in the c-Myc
recognizable structural feature in supercoiled DNA promoter: implications for drug targeting and control
molecules. Proc Natl Acad Sci USA 77, 6468–6472. of gene expression. J Med Chem 52, 2863–2874.

2294 FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS

A. Travers and G. Muskhelishvili DNA structure and function

106 Carnevali F, Caserta M & Di Mauro E (1984) 116 Sobetzko P, Travers A & Muskhelishvili G (2012)
Transitions in topological organization of supercoiled Gene order and chromosome dynamics coordinate
DNA domains as a potential regulatory mechanism. spatiotemporal gene expression during the bacterial
J Biol Chem 259, 12633–12643. growth cycle. Proc Natl Acad Sci USA 109, E49–
107 Kouzine F, Gupta A, Baranello L, Wojtowicz D, Ben- E50.
Aissa K, Liu J, Przytycka TM & Levens D (2013) 117 Sobetzko P, Glinkowska M, Travers A &
Transcription-dependent dynamic supercoiling is a Muskhelishvili G (2013) DNA thermodynamic
short-range genomic force. Nat Struct Mol Biol 20, stability and supercoil dynamics determine the gene
396–403. expression program during the bacterial growth cycle.
108 Muskhelishvili G & Travers A (2013) Integration of Mol BioSyst 9, 1643–1651.
syntactic and semantic properties of the DNA code 118 Jeong KS, Ahn J & Khodursky AB (2004) Spatial
reveals chromosomes as thermodynamic machines patterns of transcriptional activity in the chromosome
converting energy into information. Cell Mol Life Sci of Escherichia coli. Genome Biol 5, R86.
70, 4555–4567. 119 Maeshima K, Hihara S & Eltsov M (2010) Chromatin
109 Muskhelishvili G & Travers A (2013) DNA structure structure; does the 30-nm fibre exist in vivo? Curr
and bacterial nucleoid-associated proteins. In: Opin Cell Biol 22, 291–297.
Bacterial Gene Regulation and Transcriptional 120 Recouvreux P, Lavelle C, Barbi M, Conde e Silva N,
Networks, Chapter 2 (Babu MM, ed), pp. 37–52. Le Cam E, Victor JM & Viovy JL (2011) Linker
Horizon Scientific Press, Norwich, UK. histones incorporation maintains chromatin fiber
110 Travers AA & Muskhelishvili G (2013) DNA plasticity. Biophys J 100, 2726–2735.
thermodynamics shape chromosome organisation and 121 Davidson L & LaFountain JR Jr (1975) Mitosis and
topology. Biochem Soc Trans 41, 548–553. early meiosis in Tetrahymena pyriformis and the
111 Opel ML, Aeling KA, Holmes WM, Johnson RC, evolution of mitosis in the phylum Ciliophora.
Benham CJ & Hatfield GW (2004) Activation of Biosystems 7, 326–336.
transcription initiation from a stable RNA promoter 122 Woodard J, Kaneshiro E & Gorovsky MA (1972)
by a Fis protein-mediated DNA structural Cytochemical studies on the problem of
transmission mechanism. Mol Microbiol 53, 665–674. macronuclear subnuclei in Tetrahymena. Genetics 70,
112 Mavrich TN, Ioshikhes IP, Venters BJ, Jiang C, 251–260.
Tomsho LP, Qi J, Schuster SC, Albert I & Pugh BF 123 Yao MC, Choi J, Yokoyama S, Austerberry CF &
(2008) A barrier nucleosome model for statistical Yao CH (1984) DNA elimination in Tetrahymena: a
positioning of nucleosomes throughout the yeast developmental process involving extensive breakage
genome. Genome Res 18, 1073–1083. and rejoining of DNA at defined sites. Cell 36, 433–
113 Crick FHC (1968) The origin of the genetic code. 440.
J Mol Biol 38, 367–379. 124 Johnson RE, Prakash L & Prakash S (2005)
114 Travers A (2006) The evolution of the genetic code Biochemical evidence for the requirement of
revisited. Orig Life Evol Biosph 36, 549–555. Hoogsteen base pairing for replication by human
115 Sperling AS, Jeong KS, Kitada T & Grunstein M (2011) DNA polymerase iota. Proc Natl Acad Sci USA 102,
Topoisomerase II binds nucleosome-free DNA and acts 10466–10471.
redundantly with topoisomerase I to enhance 125 Hunter CA (1996) Sequence-dependent DNA
recruitment of RNA Pol II in budding yeast. Proc Natl structure. BioEssays 18, 157–162.
Acad Sci USA 108, 12693–12698.

FEBS Journal 282 (2015) 2279–2295 ª 2015 FEBS 2295