Topical JPCM Published

Home Search Collections Journals About Contact us My IOPscience
Interactions between low energy electrons and DNA: a perspective from first-principles
simulations
This content has been downloaded from IOPscience. Please scroll down to see the full text.
2017 J. Phys.: Condens. Matter 29 383001
(http://iopscience.iop.org/0953-8984/29/38/383001)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 81.103.61.126
This content was downloaded on 22/08/2017 at 10:47
Please note that terms and conditions apply.
You may also be interested in:
Going clean: structure and dynamics of peptides in the gas phase and paths to solvation
Carsten Baldauf and Mariana Rossi
Ultrafast molecules and clusters

I V Hertel and W Radloff
Applications of large-scale density functional theory in biology

Daniel J Cole and Nicholas D M Hine
Theoretical description of protein field effects on electronic excitations of biological

chromophores
Daniele Varsano, Stefano Caprasecca and Emanuele Coccia
Journal of Physics: Condensed Matter
J. Phys.: Condens. Matter 29 (2017) 383001 (29pp) https://doi.org/10.1088/1361-648X/aa79e3
Topical Review
Interactions between low energy electrons

and DNA: a perspective from first-principles
simulations
Jorge Kohanoff1 , Maeve McAllister1, Gareth A Tribello1 and Bin Gu1,2
1
Atomistic Simulation Centre, Queen’s University Belfast, Belfast BT7 1NN, United Kingdom
2
Department of Physics, Nanjing University of Information Science and Technology, Nanjing 21004,
People’s Republic of China
E-mail: [email protected]
Received 18 January 2016, revised 28 May 2017

Accepted for publication 15 June 2017
Published 16 August 2017
Abstract
DNA damage caused by irradiation has been studied for many decades. Such studies allow us to
better assess the dangers posed by radiation, and to increase the efficiency of the radiotherapies
that are used to combat cancer. A full description of the irradiation process involves multiple
size and time scales. It starts with the interaction of radiation—either photons or swift
ions—and the biological medium, which causes electronic excitation and ionisation. The two
main products of ionising radiation are thus electrons and radicals. Both of these species can
cause damage to biological molecules, in particular DNA. In the long run, this molecular
level damage can prevent cells from replicating and can hence lead to cell death. For a long
time it was assumed that the main actors in the damage process were the radicals. However,
experiments in a seminal paper by the group of Leon Sanche in 2000 showed that low-energy
electrons (LEE), such as those generated when ionising biological targets, can also cause
bond breaks in biomolecules, and strand breaks in plasmid DNA in particular (Boudaiffa et al
2000 Science 287 1658–60). These results prompted a significant amount of experimental and
theoretical work aimed at elucidating the role played by LEE in DNA damage.
In this Topical Review we provide a general overview of the problem. We discuss
experimental findings and theoretical results hand in hand with the aim of describing the
physics and chemistry that occurs during the process of radiation damage, from the initial
stages of electronic excitation, through the inelastic propagation of electrons in the medium,
the interaction of electrons with DNA, and the chemical end-point effects on DNA. A very
important aspect of this discussion is the consideration of a realistic, physiological environment.
The role played by the aqueous solution and the amino acids from the histones in chromatin
must be considered. Moreover, thermal fluctuations must be incorporated when studying these
phenomena. Hence, a special place in this Topical Review is occupied by our recent first-
principles molecular dynamics simulations that address the issue of how the environment favours
or prevents LEEs from causing damage to DNA. We finish by summarising the conclusions
achieved so far, and by suggesting a number of possible directions for further study.
Keywords: DNA damage, low energy electrons, first-principles molecular dynamics,
environmental effects
(Some figures may appear in colour only in the online journal)
1361-648X/17/383001+29$33.00 1 © 2017 IOP Publishing Ltd Printed in the UK

J. Phys.: Condens. Matter 29 (2017) 383001 Topical Review
1. Introduction
DNA was discovered in 1869 by Friedrich Miescher. It then

took about 50 years to identify the base, sugar and phosphate
components that together form a nucleotide, and to determine
that a DNA strand was simply a polymer of nucleotides con-
nected through the phosphate groups. The first suggestion that
genetic information should be encoded in a double-stranded
molecular structure was made in 1927 by Nikolai Koltzoff [2],
while the identification of this molecule as DNA was made
by Avery, MacLeod and McCarthy in 1943 [3]. The precise
determination of the structure of DNA in 1953 by Watson and
Crick [4], using x-ray diffraction data obtained by Rosalind
Franklin and Raymond Gosling, was the culmination of this
massive body of research and remains the central pillar that
underpins modern genetics. DNA contains the information
Figure 1. The four nucleobases present in DNA.
used to synthesize the proteins that perform most cellular
functions. This molecule is essential both in terms of inheri-
tance and the cellular machinery of life. It is therefore not sur-
prising that there has been a huge amount of research devoted
to understanding the structure and function of DNA under
varied circumstances.
1.1. DNA structure
The genetic information is stored in DNA as a code, which Figure 2. Phosphate and sugar components of a nucleotide.
is written into its primary structure: the sequence of nucleo-
bases. Each of the nucleobases in a DNA strand can only be
one of the four nucleic acids Thymine (T or Thy), Cytosine
(C or Cyt), Guanine (G or Gua), and Adenine (A or Ade). The
first two of these bases are aromatic heterocyclic pyrimidines,
while the other two are purines, which consist of a pyrimidine
fused to an imidazole ring, as shown in figure 1.
In the single-stranded molecule RNA, the role of thymine
is played by uracil (U), which differs from T only in that the
methyl group is replaced by a hydrogen. DNA nucleotides
are composed of one of the four nucleobases above, a sugar
component (deoxyribose) and a phosphate component (phos-
phoric acid) as shown in figure 2. The base and the sugar are
linked by the C–N glycosidic bond (A in figure 3), while the
sugar and phosphate groups are connected through the C–O
phosphodiester bond.
The backbone of DNA is built by connecting nucleotides
via the phosphate group as shown in figure 3. This introduces
two types of phosphodiester bonds, the first of which is termed
the C3-O3 bond, or simply the 3’ bond. This bond is indicated
using the letter B in figure 3 and differs from the C5-O5 bond, Figure 3. Adenosine dinucleotide showing the three components:
or 5’ bond, which is indicated using a C in figure 3, as the base, sugar and phosphate, connected via glycosidic (A) and
phosphodiester C3-O3 bonds (B). Consecutive nucleotides are
carbon atom within it is part of the sugar ring.
linked through the phosphate component (the backbone) via a
When in vivo, DNA wraps around histone proteins to form phosphodiester bond of a second type, C5-O5 (C).
chromatin, which in turn coils in the presence of the physi-
ological medium that contains water, amongst other species. during mitosis (cell division). At a genetic level, this complex
Chromatin is a macromolecular complex whose primary role is divided into chromosomes, which in turn are subdivided
is to package DNA into a more compact shape. One of the into nucleosomes containing around 150 base pairs. Figure 4
roles of this compactification is to mitigate DNA damage. shows a view of a nucleosome, i.e. a basic unit of DNA con-
However, the structure of chromatin fluctuates as the cell cycle sisting of a segment of DNA wound in sequence around eight
progresses, as it has to allow for DNA opening and replication histone protein cores, in a water background.
2
viable any longer because the genetic material is damaged,

the reproductive cycle is arrested and the biochemical material
is recycled.
If these damage processes were permanent, then there
would be an inordinate amount of cells dying continually,
and life as we know it would not be tenable. Nature, however,
has developed a sophisticated enzymatic machinery to repair
damage [5]. Interestingly, the instructions for building this
machinery are also contained in the genetic code. Therefore,
if those sections of DNA become severely damaged, then
its ability to repair damage can be compromised. Repairing
Figure 4. Nucleosome showing DNA in its physiological base excision and SSB is relatively simple, as the template
environment. DNA is wrapped around histones and solvated in remains intact. Repairing DSB is a different matter because
water. The explicit water molecules shown in the figure represent the enzyme has to connect two separate pieces of DNA and
the region affected by ionising radiation in an irradiation event. guess how to do it properly without resorting to a template.
Figure created by de Vera, after Surdutovich et al (2013 Sci. Rep. 3 The mechanisms by which enzymes repair DNA are var-
1289).
ied, and constitute a field on its own, which has been recog-
This it not yet exactly the physiological environment, nised by the award of the 2015 Nobel Prize for Chemistry to
which should take into account the interaction of the nucle- Lindahl, Modrich and Sancar. Since metabolic damage is so
osome with other fragments of the wrapped chromatin. common and independent of external sources, the enzymatic
However, it is sufficient to highlight the main components in repair machinery is likely to be tuned for such processes. It is
the DNA environment, i.e. water, RNA and proteins. In this not clear, however, how effective this machinery is at repair-
review, following most of the theoretical work done so far, we ing the damage induced by radiation. The two types of dam-
will discuss the effects of radiation on DNA in model environ age are not necessarily of the same nature, as the damaging
ments that capture the essential elements depicted in figure 4. agents are different.
1.2. DNA damage 1.3. Radiation damage of DNA
Disruption of the DNA code leads to genetic mutations that We will be concerned here with ionising radiation, i.e. ions or
can be more or less evidently expressed in observable char- photons with energies above the ionisation threshold in bio-
acteristics of the individual. Such mutations are important in logical systems, which is typically around 8.5 eV for nucleo-
both normal (e.g. evolution) and abnormal (e.g. cancerous) bases [6, 7] and 12.6 eV for water [8]. Biological matter is
biological processes. As well as mutations, which involve continually subject to ionising radiation. Some of this arises
replacing one or more base pairs with others, the structure from natural sources such as UV-light from the sun, cosmic
of DNA can also be damaged. Typical lesions of this type rays from space or from radioactive elements such as radon
include: (a) base excision, i.e. the removal of a base by cleav- that can be present in soil and rocks, our blood and bones,
age of the glycosidic bond, (b) cross-linking, i.e. the forma- and in the food and water we ingest. However, biological
tion of inter- or intra-strand chemical bonds, and (c) strand organisms are, in addition, voluntarily exposed to radiation
breaks i.e. cleavage of the 3’ or 5’ phosphodiester bond. The sources such as medical x-rays and consumer products such
most common strand breaks are single strand breaks (SSB) as luminised watches and smoke detectors [9]. On average,
in which there is a break only in one of the two strands of the the equivalent radiation dose—a measure of the biologi-
DNA double helix. These breaks are particularly relevant in cal effect—absorbed annually by humans is about 2.4 mSv.
RNA or during replication of DNA, when the double helix The official recommendation is that any exposure above the
opens for transcription purposes. A double strand break (DSB) natural background radiation should be kept below 1 mSv
is said to have occurred if both of the strands in a DNA double per year. However, background radiation is not uniform, and
helix break and if the distance between the locations of the there are regions in the world where it is significantly higher
two breaks is less than ten base pairs. than the safety limit [9]. Furthermore, in space missions at
The origin for either one of these types of damage can be the International Space Station, the typical equivalent radia-
natural, e.g. via metabolic processes. Metabolism releases, tion dose absorbed, mostly from cosmic rays, ranges between
amongst other species, reactive oxygen, nitrogen and car- 50 and 2000 mSv. Understanding the effect this radiation has
bonyl species that can attack DNA. In the human body, such upon us is of paramount importance in terms of curing and
natural oxidative DNA damage events occur at a phenomenal preventing diseases such as cancer. Furthermore, if we truly
rate of over ten thousand times per cell per day. However, want to minimise the biological effects of radiation, we need
damage can also be due to external influences, notably radia- to have an understanding of the intricacies of the processes
tion. An excessive accumulation of damage in the DNA trig- that lead to such damage. To achieve this, we need to focus
gers, amongst other possibilities, what is known as apoptotic our research on obtaining a solid understanding of how these
response, or programmed cell death. When the cell is not processes operate at the atomic and molecular level. This is
3
doubly true when delivering radiation such as photons, elec- we have to determine how fast and far these products travel,
trons from β-emitters, or ions [10] in a controlled manner for and what happens to them when they stop in the medium. In
therapeutic uses. particular, does their stopping reduce their reactivity, inacti-
Any of the components in a biological medium can be ion- vate them and prevent them from causing further damage?
ised, be they DNA, proteins, water or membranes. DNA—the This review will focus on indirect damage but, for complete-
molecule that encodes genetic information and that is respon- ness, we briefly mention some interesting aspects of direct
sible for cell replication and protein synthesis—is considered damage.
to be particularly sensitive to damage. DNA dysfunction can
cause cell death through necrosis or the relatively benign 1.3.1. Direct damage. The consequences of direct damage,
process of apoptosis (see above). Worse still, if DNA dam- i.e. the dynamics of holes in DNA, was studied intensely in
age disrupts normal cellular processes such as apoptosis, cell the late nineties and onwards. It was rapidly established that
replication can become uncontrolled and cancerous tumours holes can travel long distances [17, 18], via a hopping mech
can begin to form. anism between trapping sites. Trapped holes couple to lattice
There are two main ways radiation can cause damage to distortions forming polarons [19], and preferentially localise
DNA. One of them is directly, by ionising DNA and creat- in guanine-rich regions, although they appear to be delocal-
ing electron holes that weaken the backbone as part of a pro- ised over a few bases rather than on a single one [20]. Fur-
cess that can cause strand breaks. The other one is indirectly, thermore, there is unequivocal evidence that single and double
and is initiated by the ionisation of the surrounding medium, strand breaks occur in DNA that contains holes [21]. In this
i.e. water, proteins, etc. Any incident radiation will interact review we will not discuss holes in DNA, but it is worth men-
with these components generating secondary species, namely tioning that the discovery of high hole and electron conductiv-
electrons and radical cations, i.e. molecules with a missing ity through DNA prompted a significant amount of research
electron (or containing a hole), both of which can go on to exploring potential applications in other fields such as molec-
attack DNA. One might argue that questions of what percent ular electronics, with recent reports suggesting that this may
age of damage can be attributed to direct radiation and what become a reality in the coming years [22].
percentage can be attributed to indirect radiation are the most
fundamental in setting research priorities. 1.3.2. Indirect damage. When in vivo, DNA wraps around
It is tempting to argue that the amount of DNA in the proteins to form chromatin, which in turn coils in the presence
nucleus and the mitochondria of the cell is insignificant when of water, as shown in figure 4. Any incident radiation will thus
compared to the amount of other material in the cell. One interact with proteins and water, ionising them and generating
would thus expect indirect damage to be a much more impor- secondary species, essentially secondary electrons and reac-
tant pathway as any radiation will most likely encounter the tive radicals.
molecules in the medium rather than the DNA. While the pro- High-energy electrons, i.e. those arising from the tail of
portion of direct to indirect damage depends on the type and the distribution, interact with the medium via electron impact
energy of the radiation, experimental evidence suggests that ionisation and so generate secondary electrons while losing
the environment is responsible for approximately 66% of the a fraction of their kinetic energy. These secondary electrons
damage [11, 12]. An explanation for this large, but not over- also undergo collisions producing tertiary electrons, and so on.
whelming contribution is that, for an ionisation event to lead The overall result is a radiation cascade in which most of the
to DNA damage, it has to occur relatively close to the DNA. secondary, tertiary, and further electrons end up having ener-
An estimate based on the mean free path of electrons in water gies under ≈30 eV. The species that result from this process
suggests that only ionisation events occurring within a region are the so-called low-energy electrons (LEE) [11], and follow
between 1 and 10 nm from the DNA can contribute to the dam- an energy distribution that peaks at about 9–10 eV, as shown
age [13]. In the case of radicals, their contribution is controlled in figure 5 [23]. LEE diffuse through the medium polarising
by their diffusion properties in the medium. Typical diffusivi- it and interacting inelastically with the molecules of which
ties of OH• in water are of the order of 3 × 10−9 m2 s−1, i.e. it is composed, e.g. water, proteins, etc. As a result of these
similar to water self-diffusion [14]. These values are not too interactions LEE lose energy by transferring it to electronic
large, but in the case of ion irradiation, they can be boosted by and vibrational excitations of the molecules in the medium.
the rarefaction wave (shock wave, for heavy ions) driven by the Through this process, LEE slow down, and the majority of
ionisation of the material surrounding the ion track [15, 16]. them become thermalised, solvated electrons in a time scale
These considerations lead to the following central questions: on the order of ps. However, along their way they generate a
wealth of potentially damaging species that are described in
• How does the ionisation of the medium cause DNA
what follows.
damage?
When radiation interacts with water the following chemi-
• Do all ionisation products generated in the medium
cal reaction occurs:
interact with DNA and cause damage, or do only a frac-
tion of them do this? H2 O + radiation → H2 O•+ + e−
(1) aq ,
In answering these questions we must investigate the trans- In the above the e−aq is used to represent the hydrated elec-
port properties of the medium and its reactivity. That is to say tron and the bullet indicates that the H2O+ ion is a radical,
4
The second of these two reactions involves the radical water

cation once more and thus competes with reaction (3). As
reaction (3) is very fast and as water is much more abundant
that DNA, the majority of the radical water cations are trans-
formed into OH• radicals, and do not attack DNA directly.
Another reactive channel that avoids electron solvation
altogether is the resonant trapping of electrons by the molecu-
lar potential at specific, well-defined, positive energies. When
an electron attaches resonantly to molecule A, it forms a tran-
sient negative ion (TNI) A−∗, where ∗ signifies that this spe-
cies is in a excited state. Resonances are not strictly stationary
states in the quantum-mechanical sense. They have a lifetime
after which they undergo the reaction A−∗ → A + e−. The nar-
rower the resonance is in energy, the longer the lifetime, and
vice versa. In a TNI the electronic degrees of freedom are cou-
Figure 5. Energy distribution of secondary electrons produced by pled to the vibrational motion of the molecule. Consequently,
swift protons of various energies, and He ions in water. Notice the there is a non-negligible probability for the electronic system
mild dependence of the distribution on the energy of the projectile. to decay into a lower-lying electronic state of the molecule
Reprinted from Pimblott and La Verne [23], Copyright 2007, with while transferring energy into the vibrational motion of spe-
permission from Elsevier.
cific bonds. If the amount of energy transferred to vibrations
i.e. a molecule with a partially filled electronic shell. The two during this process is sufficiently large, this can cause bond
products of the above reaction can react with water as shown cleavage and the trapping of the electron in a bound state of
below: the molecule (at negative energy). This process is known as
dissociative electron attachment (DEA) and is known to occur
e− •
aq + H2 O → H + OH
(2)
−
when LEE interact with water molecules. A resonance at an
energy of about 7 eV [29], leads to the formation of an OH•
H2 O•+ + H2 O → H3 O+ + OH• .
(3) radical and a hydride anion via the reaction:
Reaction (2) is supposed to be slow, which is why it is widely
H2 O + e− → H2 O•− → OH• + H− .
(6)
assumed that, when electrons solvate in water and become
hydrated electrons, their reactivity drops and their potential In summary, the two main water radiolysis products are the
to induce DNA damage is reduced significantly. As a con- OH• radicals that are generated through reactions (3) and (6),
sequence, DNA damage has been traditionally assumed to and the electrons which are generated through reaction (4).
involve the radicals and not the electrons. Experimental data, Both of these radiolysis products can damage DNA and, in
however, shows that, in solution, nucleobases are indeed the long run, this damage can prevent cells from replicating
reduced by solvated electrons [24]. Very recent theoretical and thus cause cell death. For a long time it was assumed that
calculations support this notion and also estimate a time scale the radicals were the main actors in this process. However, in
between 10 and 120 fs for the reaction assuming that the e− aq 2000 experiments reported in a seminal paper by the group of
and the base are in close proximity to each other [25]. This Leon Sanche in Sherbrooke showed that LEE can also cause
reaction is then limited by the diffusion of the two species strand breaks in plasmid DNA [1]. While OH• and other radi-
relative to each other. cals remain an important source of damage, this is closer in
In contrast to reaction (2), reaction (3) is very fast and rep- nature to metabolic damage and is therefore likely to be eas-
resents the main source of hydroxyl radicals (OH•), which can ier to repair. Figure 6 summarises schematically the types of
also be formed by the dissociation of excited water molecules DNA damage due to ionising radiation.
into OH• and H•. Hydrogen radicals react with each other to In this review we explore the damage secondary electrons
form H2, while two hydroxyl radicals can react to produce can do to DNA and focus on the interactions between LEE and
hydrogen peroxide [26]. The two most important mechanisms DNA, and the manner in which such interactions lead to DNA
by which OH• radicals attack DNA are by adding themselves damage. The results presented in [1] motivated a large amount
to the π bonds of DNA bases, or by abstracting a hydrogen of both experimental [11, 30–33] and theoretical [34–36] work
atom from the sugar component. Both of these reactions are that aimed at elucidating the role played by LEE in DNA dam-
fast in contrast to reactions between OH• and phosphate [27]. age. In recent years, this research has moved on from ideal-
Therefore, strand breaks due to OH• proceed mostly via indi- ized systems, e.g. isolated DNA components in either the gas
rect mechanisms [27]. phase or a microsolvated environment, towards a more realis-
The two radiolysis products from reaction (1), i.e. H2 O•+ tic model for the environment that better represents the physi-
and e−aq can also react directly with DNA as shown below [28]: ological one. Incorporating these more accurate models for
e− •− the environment requires one to take both the aqueous solu-
aq + DNA → DNA
(4)
tion containing the ions and the presence of amino acids from
the histones in chromatin into account. Moreover, incorporat-
H2 O•+ + DNA → DNA•+ + H2 O,
(5)
ing thermal fluctuations is essential, as all processes happen at
5
Figure 6. Schematic representation of the processes involved in the interaction of radiation with biological matter. Reprinted figure with
permission from [42], Copyright (2004) by the American Physical Society.
room temperature. Understanding DNA damage under these

conditions constitutes a fundamental step towards a thorough
assessment of the dangers posed by involuntary exposure to
radiation. It also enables a finer control of radiotherapeutic
treatments of cancer, and the development of new, more effi-
cient and less harmful radiotherapies.
1.4. DNA damage by low-energy electrons
LEE generated by water radiolysis can be captured by DNA

and form the transient anion DNA•−. Much like water, DNA•−
can dissociate via DEA, using the excess electronic energy to
break specific bonds. The first experiments that detected DNA
damage at low electron energies, and that therefore probed the
DEA mechanism, were conducted by Leon Sanche’s group in
2000 [1]. Prior to this work, electrons had only been observed Figure 7. DNA damage yields caused by low energy electrons.
Peaks observed at energies in the 7–10 eV range are ascribed to
to cause DNA damage when they had energies above 30 eV. resonant behaviour, leading to dissociative electron attachment.
This was one of the reasons why OH• radicals were believed These resonances fall in the same region of those in water. Reprinted
to be the key agent in indirect DNA damage [26]. The observa- from Luo et al [39], Copyright (2014), AIP Publishing LLC.
tion that LEE could cause damage is particularly significant,
because electrons are as or more abundant in irradiated cell reported in [1]. This figure shows that a large number of
conditions than radicals [37]. In Sanche’s experiments, plasmid lesions are observable at energies lower than the ionisation
DNA was irradiated with electrons having well-defined inci- potential of DNA, which is between 7.5 and 10 eV across the
dent energies that ranged from 3 to 20 eV. The number of DNA four nucleobases [40]. This suggests a low energy mechanism
strand breaks per incident electron was measured using the is at play. The authors argued that this mechanism was dis-
gel electrophoresis technique. This technique works because sociative electron attachment leading to the cleavage of the
undamaged DNA, DNA containing SSB, and DNA that con- phosphodiester bond—a process which manifests itself as a
tains DSB, have different shapes and sizes, and hence travel at strand break. This hypothesis was soon supported by Barrios
different speed through the gel under the influence of an applied et al [41] who showed, using first-principles calculations, that
electric field. This allows for a quantitative evaluation of the dif- the excess electron is captured in a π ∗ orbital located in the
ferent types of damage [38]. A main finding is that the amount nucleobase, and transferred remotely to the phosphodiester
of DNA damage does not grow monotonically with energy, but bond due to the electron-vibration coupling. A few years later
exhibits an interesting structure, as shown in figure 7. Sanche’s group extended their experiments to lower ener-
Figure 7 reports the yields of crosslinks, SSB, DSB, and gies and showed that strand breaks can occur even when the
loss of supercoiled configuration, as a function of the inci- electrons have very low energies, between 0 and 4 eV [42].
dent electron energy, [39] and expands on the original results They observed strand break yield peaks at 1 eV and 2.5 eV
6
Figure 9. Thymine base with ring atoms numbered. When a 1 eV

electron forms a TNI with it, cleavage occurs at the N1-H bond.
At 1.8 eV, cleavage is observed at the N3-H bond [51]. Reprinted
from [63], with the permission of AIP Publishing.
of DNA. At higher energies the TNI formed are Feshbach

resonances.
Figure 8. The peak/valley distribution of SSB is still observed at
even lower energies (0–4 eV), with peaks observed at 0.8 eV and 1.4.2. Dissociative electron attachment (DEA) in nucleo-
2.2 eV. DSBs are not observed at such low energies. Reprinted with bases. The pathway via which a TNI dissociates depends on
permission from Martin et al [42]. Copyright 2004 by the American the resonant energy at which it is formed. In all these path-
Physical Society.
ways, the excess energy of the attached electron transfers
(See figure 8) and supported these observations with cross through the molecule causing the dissociation of a bond that
section calculations that further reinforced the DEA picture. may well be quite distant from the region where the electron
attaches. Crossed-beam collision experiments on each of the
1.4.1. Electronic resonances in DNA subunits. The connec- four nucleobases found that low-energy resonances dissoci-
tion between resonant energies in DNA and capture in the ated via the loss of a neutral hydrogen atom [50], leaving a
nucleobases was corroborated by comparing to earlier gas dehydrogenated negative ion [Thy-H]–. Extensive studies
phase experiments by Huels et al [43], in which a molecular of the Thymine nucleobase [51] found that, at typical shape
beam of the DNA bases was made to interact with an electron resonance energies (0–3 eV), there was an interesting energy-
beam. An analysis carried out by mass spectrometry to iden- dependent bond selectivity. For example, N1-H cleavage
tify ion signals clearly revealed two low-energy resonances, resulted from a 1 eV resonance, while at 1.8 eV, cleavage was
one at around 1 eV, and the other in the range 6–12 eV. There observed in the N3-H bond (see figure 9 for the locations of
are thus two major types of transient negative ions (TNI) these bonds). When the TNI was formed at the higher energy
formed via LEE attachment to DNA subunits (nucleobases, resonance (6–12 eV), the prominent reaction was the release
sugar or phosphate components) that can lead to damage [33]: of a hydride H− ion, leaving a neutral [Thy-H]• radical. This
reaction was observed to extend over the whole resonance
(i) Shape resonances, in which the electron temporarily region [51].
occupies a previously empty orbital of a DNA subunit The transient negative ion is a challenging species to
and is trapped by an angular momentum barrier. At low investigate, precisely because it is metastable and relatively
energies the TNI will exist for long enough for DEA to short-lived. One way to identify them experimentally uses
occur [44] but, at higher energies, the quasi-bound elec- crossed-beam scattering experiments in which the energy of
tron will tunnel so rapidly that it will escape the molecule the incident electron beam is scanned over the desired energy
before DEA can take place [45]. region. Peaks in the cross section correspond to DEA reso-
(ii) Core-excited Feshbach resonances, in which the electron nances, and can be measured via mass spectrometry of the
is captured in an electronically excited state of the DNA fragments detected in the exit channel. In DNA, such peaks
subunit. The molecule will eventually decay in to the are observed in the region 0–10 eV [53].
parent state, but Feshbach resonances are generally quite From the theoretical side, several methodologies have
sharp and hence longer-lived than lower energy shape been developed to compute scattering cross sections. Most of
resonance. Furthermore, they carry a larger damaging the work on biomolecules and DNA subunits has been carried
potential because a larger energy is available for transfer out using R-matrix theory. In R-matrix theory, the scatter-
to vibrations [34]. ing center, e.g. a molecule, is enclosed in a spherical region
Gas phase experiments [43, 46–52] revealed that at low inside which the electronic structure is determined to a high
electronic excitation energies (up to 3–4 eV), the TNI that are accuracy, typically at the configuration interaction level. The
most likely to form are shape resonances of the basic subunits internal wave functions are then matched at the surface of the
7
sphere with external incoming and outgoing waves, which are

expanded in partial waves. The enforcement of such bound-
ary conditions, similarly to the case of scattering by a one-
dimensional potential well, determines the scattering cross
section as a function of energy. This methodology exhibits
plenty of numerical subtleties. One of the most important
ones is a need for (pseudo-)states in the continuum that cor-
respond to scattering, rather than to bound states of the poten-
tial. A second, very important choice is the basis set that is
used to represent the wave functions inside the sphere. For
atoms it is more or less straightforward, but for molecules one
is faced with the question of whether to use a single-center or
a multi-centre expansion. Both have their advantages and dis-
advantages. A thorough review of R-matrix theory and appli-
cations to molecular resonances can be found in [54]. This
formalism considers fixed nuclei, which is not appropriate to
the computation of DEA cross sections, where the electronic Figure 10. Cross-section for DEA of uracil at N1-H comparing
and nuclear dynamics are coupled. Extensions of R-matrix R-matrix theory results with experiment. The height and shape of
theory that include nuclear dynamics have been proposed the calculated cross-section (black line) is in good agreement with
experiment (blue line). The experimental peak however occurs at
by Fabrikant [55] and Domcke [56], amongst others. The an energy 0.27 eV lower than the R-matrix peak. Reprinted with
quasi-classical approach of Fabrikant uses ab initio poten- permission from Gallup and Fabrikant [60]. Copyright 2011 by the
tial energy surfaces (PES) and resonance widths as a func- American Physical Society.
tion of nuclear geometry as input data for the computation of
DEA cross sections. This formalism is suited for the excita- value, 1.1 eV, but the overall shape of the curve is very well
tion of a single vibrational degree of freedom, which works reproduced.
well for diatomic molecules, but in a more general context, it These are gas-phase results. It is not quite clear what are
requires the freezing of all but one vibrational degree of free- the implications for more representative environments, e.g.
dom. More general schemes that use multi-dimensional PES when the base is solvated in water. In order to start answer-
have also been proposed [57]. The theoretical description of ing this question, Smyth et al extended the previous work by
the coupled electron-nuclear dynamics as it occurs in DEA supplementing R-matrix calculations with multiple scattering
remains a challenge. Some alternative ideas will be discussed theory to take the effect of microsolvation, i.e. the presence of
in section 2. a small number of water molecules hydrogen-bonded to ura-
Caron and Sanche investigated low energy resonances in cil, on the DEA process into account [63]. A first finding was
DNA using R-matrix calculations for the atomic centres com- that the N1-H resonance shifted towards lower energies due to
bined with a multiple scattering approach to take into account the stabilising role of the environment. More importantly, the
the molecular environment [58]. In this way they calculated width of the resonance peak reduced, thus increasing the life-
cross sections for large, model helical DNA structures, and time of the resonance and enhancing the DEA cross section by
predicted resonances that were below 15 eV and which would a factor of six, as shown in figure 11. This suggests that dis-
contribute to an explanation for the stand breaks observed in sociation of [Thy]– is more likely in an aqueous environment
experiments. A more accurate determination was provided than it is in the gas phase.
later using a similar approach, but using collisional data for
DNA subunits instead of atomic data [59]. 1.4.3. DEA in other DNA subunits: sugar, phosphate, nucleo-
During the DEA reaction the electron decays into a bound sides and nucleotides. So far we have only reviewed work
state while releasing its excess energy and causing dissocia- on the DEA of nucleobases. We have established that low-
tion. The resonant energies represent the excess energy that energy electrons (0–10 eV) are likely to be captured resonantly
will be used to excite bond vibrations and thus make bonds leading to dissociation via cleavage of the N–H bonds. This
prone to cleavage. DEA cross sections can be calculated phenomenon, however, is not directly relevant to the strand
using R-matrix theory, as shown by Gallup and Fabrikant breaks observed in Sanche’s initial work on plasmid DNA [1,
[60], who investigated the dissociation of uracil. Using a 42] as strand breaks are associated with the cleavage of the
finite element discrete model [61], they calculated the PES phosphodiester C–O bonds between sugar and phosphate that
for the TNI corresponding to LUMO of a1 symmetry. This hold the DNA strand together—the bonds labelled B (3’) and
resonance is very broad at the equilibrium nuclear configura- C (5’) in figure 3. Instead, the cleavage of bond A, i.e. the
tion, but becomes narrower when the N1-H bond is stretched, glycosidic C–N bond, leads to base excision.
eventually leading to the formation of a stable [U-H]– ion. The The sugar moiety in the gas phase also displays resonant
nuclear dynamics was then included into R matrix theory, and attachment properties, but it is less likely to attach an electron
a pronounced vibrational Feshbach resonance [62] was found and decay than a nucleobase. Studies of electron attachment
near the DEA threshold, see figure 10. The position of this to an isolated sugar found two resonant features. One at 0 eV,
resonance, 1.4 eV, is somewhat above the experimental peak and a broader resonance at 6–9 eV. These resonances decay
8
the bonds which can break in longer strands of DNA are the
C3-O3 (B) or C5-O5 (C) phosphodiester or the C–N (A) gly-
cosidic bonds shown in figure 3, above. Soon after the original
low-energy electron experiments by Sanche’s group, Barrios
et al [41] proposed the following mechanism for SSB. A low-
energy electron (≈1 eV) attaches to a π ∗orbital of a DNA base
to form a shape-resonance. This anion then undergoes a sugar-
phosphate C–O bond rupture over a small barrier to produce
a SSB. These authors also suggested that solvation should
play a crucial role in the rate of SSB formation by stabilising
the anion, which otherwise will be very short-lived against
auto-detachment.
This mechanism has been revisited many times, e.g. [72,
73], but mostly from the point of view of ground state elec-
tronic structure calculations, which will be reviewed later.
Full scattering R-matrix calculations have been performed for
systems as large as single DNA nucleobases in vacuum [74,
Figure 11. DEA cross section for uracil: isolated (black solid line) 75], and microsolvated by up 5 water molecules [76]. This is
and embedded in the (H2O)5 cluster. The location of the resonance practically the state-of-the-art for the calculation of electronic
is shifted by ∆E = −0.326 eV. The dotted red curve is the cross resonance cross-sections, while for DEA probably the largest
section calculated with inclusion of scattering by all molecules
in the cluster, while the dashed blue curve considers scattering system studied is the one in [63]. Theoretical and experimental
from water molecules. Reprinted from Smyth et al [63], with the progress in DEA has been reviewed very recently [77].
permission of AIP Publishing. In what follows we will discuss the post-DEA behaviour of
DNA subunits after the excess energy of the attached electron
via the loss of one or more neutral water molecules and CH2 O
has been transferred to vibrations. The aim is to predict the
molecules [64–66] and the associated reactions are more com-
likelihood of the reactions that could result from TNI decay
plex than the dehydrogenation of the nucleobases. Quantum
in different environments, from gas-phase to full solvation,
dynamics scattering calculations [65], confirmed that shape
as opposed to thermally-activated, electronic ground-state
resonances in the ribose should not exist at 0 eV but will be
reactions.
produced between 7–10 eV. The dissociation that is observed
at 0 eV is thus still unexplained. However, when the sugar is
bound to the nucleobase, as in a nucleoside, DEA is likely to 2. Modelling the dynamics of DEA
occur. This was observed by Ptasinska et al, who performed
thermal evaporation experiments on Thymidine [67, 68]. At The process of energy transfer from electronic excited states
energies below 3 eV, the molecule exhibited two resonances: to vibrational motion is a typical example of non-adiabatic
the cleavage of the glycosidic bond at 1.2 eV, and a shape reso- dynamics. Several methodologies have been developed during
nance at 1.8 eV on the nucleobase, which resulted in a release the past decades to describe such phenomena. It is not the aim
of the N3-H hydrogen. of this review to discuss them in detail, but since this is one
The investigation of the isolated phosphate unit as it of the methodological frontiers, we will mention here some of
appears in a DNA nucleotide is experimentally difficult. these approaches.
One possibility is to use phosphoric acid derivatives bound First of all, the description of the electronic subsystem
to hydrocarbons as was done by Konig et al [69]. They used requires one to account, reasonably accurately, for excited
(C4 H9 O)2 POOH because of its closeness in structure to DNA. states. This rules out immediately approaches such as stand-
In terms of DNA damage, the most significant reaction was ard DFT. While originally formulated as a ground state theory
the C–O bond cleavage, which is equivalent to a strand break. DFT, in principle, also allows for the determination of excited
This reaction took place in two broad resonances at 2–4 eV states. Formally, the electronic density determines univocally
and 7–10 eV [70]. Electron attachment to the sugar-phosphate the external potential that leads to it, and this potential in turn
complex was also shown to lead to dissociation in the gas- determines the full many-body wave function, which includes
phase by Bald et al, who used laser-induced acoustic des- excited states. Determining the wavefunction in this way
orption techniques [71]. This technique was used to transfer is, however, not a practical recipe. Instead, it has been cus-
intact neutral biomolecules into the gas phase where they tomary to identify the energy difference between empty and
can interact with an electron beam. In this study, the ribose- occupied Kohn–Sham, one-electron eigenstates as excitation
5’-phosphate formed a TNI at 0 eV which resulted in cleavage energies of the many-electron system. The most prominent
of either the C–O bonds. example is that of the HOMO-LUMO gap in molecules, or
We can thus conclude that, of the individual building similarly the valence-conduction band gap in bulk semicon-
blocks of the DNA nucleotide, the nucleobase is the most ductors and insulators. This interpretation as excitation energy
sensitive to low-energy electron attachment. The base-sugar is often criticised by stating that the Kohn–Sham eigenstates
and sugar-phosphate components are sensitive to DEA and are ‘meaningless’ as these states correspond to a fictitious,
9
non-interacting reference system. However, it has been shown hamiltonians are good for developing new methodologies [81],
that this is a well-defined approximation to true excitation while tight-binding models, especially the so-called density-
energies [78], and its quality depends on the exchange-cor- functional tight-binding (DFTB) approach, has proved useful
relation functional used. In addition, it is nowadays possible for describing electron dynamics in large systems [82, 83].
to compute electronic excitation energies in terms of one- Introducing the nuclear dynamics also requires approx
electron eigenvalues to a very good accuracy using hybrid imations for the electron-nuclear correlation. The simplest
functionals that combine standard DFT approximations, such one is to evaluate the forces on the atoms in correspondence
as the GGA, with Hartree–Fock exchange [79]. Quantum with the time-evolving electronic density. In this approach the
chemical methods also provide good estimates of excitations dynamics of the electronic density follows the von Neumann
energies, but not at the Hartree–Fock (single-determinant) equation which, within TDDFT, is equivalent to integrating
level. The lowest acceptable level is configuration interaction the time-dependent Kohn–Sham equations. This is called the
including single excitations (CIS). Ehrenfest approximation, and it is essentially a mean-field
An alternative to the methods described in the previous approach in which the individual nuclei see the electrons
paragraph is to use time-dependent DFT (TDDFT), either in through their density, but not individually. While electronic
the frequency domain via linear response, or in real time. For excitation processes are captured correctly by this approach
finite systems such as molecules, standard TDDFT approx [84], the mean-field character of Ehrenfest distorts the char-
imations such as the adiabatic GGA (AGGA), which is local acteristics of energy transfer from electrons to phonons, e.g.
in time and semi-local in space, corrects the DFT excitation Ehrenfest cannot describe properly ubiquitous phenomena
energies quite well, unless they are of the charge-transfer type. such as Joule heating [85] or the thermalisation between elec-
For extended systems, however, the correction is diluted and tronic and nuclear degrees of freedom [81]. A second limi-
becomes ineffective. Interestingly, TDDFT has been demon- tation is related to the exchange-correlation approximations
strated in the computation of cross sections for electron-atom used in TDDFT. In most cases, these are semi-local in space
scattering [80]. This may be an area to explore in the near and local in time. Therefore, memory effects in the electronic
future, as TDDFT allows for the study of much larger systems evolution are ignored, which leads to another whole host of
than the quantum chemical approaches typically used in tradi- pathologies, mainly related to the lack of electronic decoher-
tional scattering theory (CI). ence (no electron–electron collisions term) [86]. In addition,
The second aspect to take into account is the correlated incoherent electron–phonon scattering by nuclear vibrations
motion of electrons and nuclei. The combination of ground- (phonons), which are not explicitly represented in classical
state DFT for electrons with Newtonian dynamics for the molecular dynamics, are not accounted for in Ehrenfest simu-
(classical) nuclei has been an extremely valuable tool in the lations [87]. While Ehrenfest dynamics is straightforward to
study of materials and biological systems. When using this implement and computationally efficient, it is not trivial to go
technique the electrons are maintained in the ground state beyond it.
corresponding to the instantaneous nuclear configuration, To improve on the electron-nuclear correlation, a possibil-
and hence have no independent dynamics. This methodology ity that has attracted interest is surface hopping [88]. In this
relies on the separation of electronic and nuclear time scales, method the forces on the nuclei are determined from single
which is rooted in the Born–Oppenheimer approximation. electronic potential energy surfaces (PES) but hops between
The numerical integration of this scheme receives the name of surfaces are allowed in order to include non-adiabatic effects.
Born–Oppenheimer molecular dynamics (BOMD), although Surface hopping works reasonably well when non-adiabatic
it is sometimes called first-principles MD (FPMD), quantum transitions occur between a small number of PES, but not for
MD (QMD) or ab initio MD (AIMD). The so-called Car– a dense manifold of excited states. Perhaps the most sophis-
Parrinello molecular dynamics (CPMD) also falls within this ticated way to go beyond Ehrenfest in a controlled manner is
class of methods, but it is implemented in a way that resorts the correlated electron–ion dynamics approach (CEID) [85].
to a fictitious dynamics for the electronic degrees of freedom CEID relies on expansions of the von Neumann equation for
[79]. In this sense, it is a precursor to methods that treat the the electron-nuclear system, with different formulations pro-
electronic wave function or density as a dynamical variable, posed in the limits of weak and strong electron-nuclear cou-
with the aim of re-introducing a proper independent electronic pling. Although CEID is still rather young and costly, it has
dynamics. The cost of this approach is to reduce significantly emerged as a powerful tool for problems in which the trans-
the integration time step, typically by a factor of thousand, fer of energy between electrons and nuclei is crucial. A cost-
thus compromising the ability to follow the nuclear dynam- effective alternative that limits the nuclear motion to harmonic
ics in the same simulation, unless the system size is reduced phonons, named Effective CEID, has been proposed recently
significantly or inexpensive approximations to the electronic [81], and applied to inelastic electron transport in water chains
structure are used, such as simple model hamiltonians. Of [89]. It is important to remark that not one of these methods
course, it is possible to improve this description by resort- is presently mature enough to allow for the dynamical simula-
ing to higher-level approaches such as tight-binding (TDTB), tion of the dissociative electron attachment process. However,
quantum chemistry methods (e.g. TDHF), or TDDFT. Model it is envisaged that this will become possible in the near future.
10
3. Dynamics ensuing DEA: from the gas to the

condensed phase
In the absence of an adequate methodology for describing

the dynamics of DEA and the subsequent evolution, we have
resorted to the following approach: we assume that the DEA
process occurs over a short time scale and that once this pro-
cess is concluded the excess electron is trapped in a bound
state of a stable anion. In other words, after DEA the DNA and
electron form a ground state anion. We assume that the excess
electronic energy is transferred into a specific vibrational mode
and as such we introduce an instantaneous kick in the kinetic
energy by modifying the velocities of the atoms involved in
that vibrational normal mode. We then follow the dynamics
of the ensuing evolution using BOMD, i.e. a combination of
DFT for ground state electrons with molecular dynamics for Figure 12. Thymine nucleobase. The N–H bond that receives the
excess energy from the excited electron is circled in red.
the nuclei. In the results presented below, we calculated the
electronic structure using the ab initio quantum module quick-
step (QS) of the open source computer code CP2K [90]. The
1 mA
ηA2 + mA vA − vB · µ̂ ηA = ∆E,
(0) (0)
Kohn–Sham orbitals and electronic density were represented mA 1 +
2 mB
using the Gaussian and plane waves method (GPW). For the (11)
exchange-correlation functional we used PBE supplemented with ηB = −mA ηA /mB. Typical values of ∆E are much larger
with a van der Waals dispersion correction. GTH pseudopoten- than thermal energies, so in principle once could disregard the
tials were used in conjunction with the TZVP-GTH basis set. last term in (11), and hence the new velocities will simply be,
The motion of the atomic system was followed using MD, with
a timestep of 0.5 fs. In our simulations, the specific vibrational 2∆E 2∆E
modes that were excited in the way described above, corre- ηA = ; ηB = − .
mA (1 + mA /mB ) mB (1 + mB /mA )
sponded to the vibration of the N–H bond in the nucleobase (12)
and the vibration of the C–O bond in the nucleotide.
To increase the vibrational energy of a molecular bond, we
modified instantaneously the velocities, and hence the kinetic 3.1. Nucleobase
energies, of the constituent atoms. For an A–B bond, the new We now present results for the dynamics following DEA in
velocities of atoms A and B are given by: a nucleobase. In this study we examined Thymine, in the gas
(0)
vA = vA + ηA µ̂
(7) ;
(0)
vB = vB + ηB µ̂ phase, in a microsolvated environment, and finally in the con-
densed phase. A vibrational analysis showed that the normal
(0) (0)
where vA and vB are the new velocities and vA and vB are the mode vibrating with the highest frequency was the N–H bond
unmodified velocity vectors of atoms A and B, respectively. highlighted in figure 12 (circled in red). We thus proceeded to
µ̂ is the director of the bond connecting atom A to atom B, investigate the effect of injecting energy into the N–H bond
which is calculated using: shown in figure 12 in the manner described above i.e. by aug-
rA − rB menting the velocity of the H-atom and, to conserve momen-
(8) µ̂ = tum, also augmenting that of the N-atom. This was done after
|rA − rB |
a period of equilibration at room temperature. Simultaneously
where rA,B are the position vectors of atoms A and B before with the kick, we added an electron to the system, which went
the kick. The quantity ηA,B is the extra velocity that will be on to occupy the LUMO of the neutral system, and turned
added to each atom. This quantity is calculated by identifying it into the so-called singly-occupied molecular orbital, or
the change in kinetic energy with the electronic energy lost by SOMO, of the anion.
decaying from the resonant state to the LUMO (∆E )
1 1 3.1.1. Gas phase. In the gas-phase we expected the bond to
mA ηA2 + mB ηB2 + mA ηA vA · µ̂ + mB ηB vB · µ̂ = ∆E,
(0) (0)
break when the N–H bond received an additional energy larger
2 2
(9) than the dissociation energy, which in this case equals 1.67 eV
together with conservation of momentum: when it is calculated at the same theory level and using the
same basis set as was used for the dynamical simulation. This
mA ηA + mB ηB = 0.
(10) value corresponds to the dissociation of Thymine into [Thy-
In these expressions mA and mB are the masses of atoms A and H]– + H, with H leaving as a neutral species. The calculated
B respectively. Combining (9) and (10) we obtain the follow- dissociation energy for the reaction [Thy-H] + H– is 5.87 eV,
ing quadratic equation for ηA: so we do not expect to see the hydride ion at low energies.
11
with the Thymine at all possible positions involving N and

O-atoms (C-atoms do not make significant hydrogen bonds),
and were taken from previous studies [63, 91]. The water mol-
ecules in this system provide a representation for the first sol-
vation shell of Thymine.
In order to estimate the minimum energy required to break
the same N–H bond as in the gas phase, we used the method
above to calculate new velocities for the N and H atoms, with
initial positions and velocities taken from a previous equi-
libration run. In contrast to the gas phase results, an excess
energy of 2 eV was not sufficient to break the bond under
microsolvated conditions. The H atom temporarily detached
from the nucleobase and interacted with nearby water mole-
cules, but eventually the bond reformed and the excess energy
was redistributed among the H-bonded water molecule and
the base. When an excess energy of 3 eV was introduced, the
H atom had enough energy to reach and collide with a nearby
water molecule. In this case we observed a concerted proton
transfer process at play. The ‘kicked’ H-atom was transferred
Figure 13. Thymine microsolvated by five water molecules. Two as a proton to the H-bonded water molecule, while one of
types of hydrogen bonds between water and Thy are present: the water H-atoms proceeded to leave the system as a neutral
O–H ⋯ O, with water donating the proton, and O ⋯ H–N,
atom. The net result was a dehydrogenated Thymine anion
with water accepting the proton. Dots indicate hydrogen bonds.
[Thy-H]–, five intact water molecules, and a reversal in the
These dissociation energies are consistent with experimental direction of the H-bond from N–H ⋯ O to N ⋯ H–O. This
data [46–49, 51] which shows that hydrogen atoms are sequence of events is shown in figure 14.
released when Thymine interacts with electrons with energies This result implies that dissociation, even in a small amount
less than 3 eV, and with results of Ptasinska et al [50], who of water, requires more energy than in gas phase. It would,
showed that hydride anions are released when Thymine inter- therefore, seem that some of the energy in excess of the
acts with electrons with energies between 5 and 13 eV. We ran 1.67 eV required for bond cleavage is used in other processes
dynamical simulations injecting kinetic energies of 1 eV and in the microsolvated environment. It is thus tempting to con-
2 eV into the N–H bond. In the first case we observed that clude that the energy required to dissociate the bond increases
the H atom made a large excursion but then returned reform- as the number of solvation waters increases. Interestingly,
ing the bond and redistributing the excess energy amongst recent experiments in microhydrated uracil and Thymine sup-
the other vibrational modes. For the 2 eV injection, the bond port the notion that caging stabilises the anions and supresses
broke and never reformed. We also observed dissociation of fragmentation [92].
the bond in a simulation in which 1.7 eV of additional kinetic
energy was added, which is just above the dissociation thresh- 3.1.3. Condensed phase. In this work condensed phase
old in the gas phase. systems have been modelled using periodic boundary condi-
Dissociation energy calculations are possible in the gas tions in all directions. This allows us to model fully-solvated
phase but not in the condensed phase. In the condensed phase molecules using a relatively small number of atoms and to
the dissociation fragments are difficult to define because the remove surface effects. The results obtained in this way are
dissociated system could assume a wide variety of different the most comparable with what might occur in cell-like condi-
configurations. In the condensed phase, we would need to cal- tions, although the concentration of nucleobase may be some-
culate the free energy for dissociation, which requires a large what larger than the typical concentrations in cells because
amount of statistics to justify convergence. We thus proceeded of the high ratio between the number of solute and solvent
in a more empirical way and studied different excess energies molecules. Our condensed phase model consisted of a single
and analysed whether or not the N–H bond dissociated. The Thymine molecule in a bath of 30 water molecules.
effect of a water environment on DEA was investigated in two During the equilibration run, the system fluctuated from
steps. We first looked at microsolvation of Thymine by five an H-bonded configuration in which the distance between
water molecules, and then we examined the Thymine-water the H of the N–H group and the oxgyen of the nearest mol-
system in the condensed phase. ecule was between 1.5–2.6 Å [93], and configurations in
which this hydrogen bond was not present. An example
3.1.2. Microsolvation. The microsolvated system consists of of a H-bonded configuration, with the closest water mol-
a Thymine base surrounded by five explicit water molecules ecule and the nucleobase are highlighted, is shown in fig-
(See figure 13). These water molecules make hydrogen bonds ure 15. Typically, fluctuations between H-bonded and
12
Figure 14. The sequence of events when the H-atom interacts with the water molecule to cause its dissociation. Snapshots are ordered
clockwise and numbered, starting from the upper left corner, and the sequence indicated with arrows. Dotted lines indicate hydrogen bonds.
environment we simulated the post-DEA dynamics in both

H-bonded and non-H-bonded configurations of the condensed
phase system. To estimate the dissociation energy for the thy-
mine anion in the condensed phase, we chose 15 H-bonded
configurations and eight non-H-bonded configurations. We
prepared the initial velocities of the N and H atoms for each
of these configurations using the method explained above. We
first studied their behaviour when an excess energy of 1 eV
was introduced and gradually increased this energy, only
stopping once the nucleobase dissociated during the subse-
quent simulation. The energy at which we stopped was a first
estimate of the dissociation energy. We also tested energies
above our first estimate in order to ensure that dissociation
still occurs. Of the 15 H-bonded configurations we examined,
we found that 12 required an energy greater than 3 eV to dis-
sociate. Of these, six configurations required more than 5 eV
to dissociate, but all configurations dissociated with an energy
Figure 15. A snapshot from a trajectory for the condensed phase
system in which the H-bond is present. In this frame, the H-bond of less than 10 eV. From the eight non-H-bonded configura-
length is 1.554 Å. tions, seven of these dissociated with an energy of 3 eV or
less. The remaining configuration required an energy between
non-H-bonded situations are seen multiple times in only 3 and 5 eV.
5 ps of simulation. On top of this we observed that the iden- We focused on the simulations in which 3 eV of excess
tity of the water molecule that was closest to the Thymine energy was introduced. We observe two different scenarios in
changed during the simulation. these simulations. When the nucleobase is H-bonded the N–H
The microsolvation results suggest that H-bonding between bond does not break and the excess energy is transferred to
the water and the nucleobase should stabilise the anion and the environment. When the nucleobase is not H-bonded, the
thus increase the energy required to drive the dissociation of bond breaks with ease and the hydrogen leaves together with
the N–H bond. This is consistent with evidence from Kumar the excess electron in the form of a neutral H-atom, which
et al that dissociation barriers are higher when a hydrogen- interacts very weakly with the water environment and hence
bond network exists between DNA and the surrounding water diffuses away rapidly. The obvious question is, why would
environment [94, 95]. On top of this we will show in sec- dissociation require more energy if there is a hydrogen bond
tion 4.1 of this review that hydrogen bonds between all of the present? A possibility is that energy is more efficiently chan-
DNA nucleotides and their aqueous environments increase nelled through the hydrogen bond and delivered to other parts
the free energy barriers for bond breaking. To examine what of the system. To investigate this, we calculated the kinetic
happens to the barrier to dissociation in a fully solvated energy of the N–H bond, the kinetic energy of the remaining
13
breaking of this bond. In the non-hydrogen bonded case, the

kinetic energy of the nearest water molecule fluctuates in a
similar way to the equilibrated system but there are no indica-
tions of a transfer of energy away from the N–H bond.
We can, therefore, conclude that the hydrogen bond serves
both as a channel to transfer energy to the surrounding solu-
tion and as an obstacle that reflects energy back into the nucle-
obase. The hydrogen bond has a caging effect on the dynamics
of the hydrogen atom that forms part of the N–H bond. This
process of caging channels the kinetic energy out of this bond
and into the surroundings and is thus what prevents it from
breaking. The presence or absence of this hydrogen bond is
thus crucial for understanding whether or not the nucleobase
will dissociate via DEA in solution. It is known that the thy-
mine anion is a weak base, with a pKa = 6.9 [96], which
Figure 16. The change in kinetic energy of the N–H bond. The suggests that there will be similar fractions of H-bonded and
black (red) line corresponds to a H-bonded (non-H-bonded) initial non-H-bonded base anions. A more precise determination of
configuration. The blue line shows the change in the kinetic energy the propensity for this hydrogen bond to form requires the use
in simulations when no additional kinetic energy was injected into
the N–H bond. of molecular simulation techniques (work in progress).
atoms in the nucleobase and the kinetic energy of the water 3.2. Nucleotide
molecule that was closest to the hydrogen of the N–H bond,
separately. For each of these groups of atoms, we then evalu- The previous section gives an insight into the effect of sol-
ated the difference between the kinetic energy before and after vation on post-DEA dynamics, and shows that nucleobases
the ‘kick’. This quantity provides a measure of how much the such as Thymine are more stable against DEA when they are
kinetic energy deviates from its unperturbed, equilibrium in solution than they are in gas phase because of hydrogen
value. bonding and caging effects. This, however, is not directly rel-
Figure 16 shows the running average for the N–H bond evant to DNA strand breaks as any base in DNA is attached
kinetic energy as a function of time. The black line is for an to a sugar and a phosphate, and H-boned to another base.
H-bonded initial configuration, while the red line is the result The next step, then, is to move to the simplest model that
in the non-H-bonded case. The blue line is a reference that contains a C–O bond like those in a phosphodiester bond.
was obtained by simulating the dynamics without the addition Deoxythymidine monophosphate (dTMP) is a nucleotide
of any kinetic energy. It is clear that the presence of the hydro- composed of a Thymidine nucleoside and a 5’ phosphate
gen bond influences the system’s return to equilibrium as the component. Several studies have shown that, instead of break-
kinetic energy of the N–H bond decays much more slowly ing at the 5’ C–O bond, strand breaks are significantly more
when the hydrogen bond is present than when it is not. When likely to occur at a 3’ bond. Therefore, instead of studying the
the hydrogen bond is absent (red line), the N–H bond breaks standard dTMP, we have focused on the variant containing a
early in the simulation. A significant fraction of the excess 3’ bond that is shown in figure 17. We are interested in the dis-
kinetic energy is converted into the potential energy required sociation energies and dissociation products associated with
to break the bond, while the remainder is redistributed. When breaking the 3’ bond and in any differences that occur when
the hydrogen bond was present the trajectories showed that the nucleotide is in different environments. We can estimate
the initial 3 eV of excess vibrational energy caused the H atom the energy required to break the C–O bond in the gas phase
to detach but that it rapidly reattached to the nitrogen. The and then compare this energy to that required when an aque-
fact that the kinetic energy of the bond decays more slowly in ous environment is introduced.
this second, H-bonded scenario, is consistent with the obser-
vations. The kinetic energy of the bond does eventually relax 3.2.1. Gas phase. We expect the C–O bond to break when the
down and evolve towards equipartition but the question of kinetic energy injected into this bond exceeds the dissociation
where the excess energy goes still remains. energy as this is what was seen for the nucleobase. The nucle-
In the H-bonded case, we observed that most of the excess otide anion can break by either releasing a Thymidine anion
energy is redistributed amongst the vibrational modes of the (dThd–) and a phosphoric acid radical (H2O4P), or by releas-
nucleobase, although a fraction of this energy takes brief ing a Thymidine radical (dThd) and a stable phosphoric acid
excursions to the nearest water molecule. Hydrogen bonds are anion (H2O4P–). Calculations performed at the PBE/TZVP-
less efficient than covalent bonds in redistributing vibrational MOLOPT level indicate that the first of these two reactions
energy, but over a longer time scale the whole system should requires an energy of 1.23 eV, while the second one only
approach thermal equilibrium. However, environmental fluc- requires 0.08 eV. Therefore, the most likely dissociation prod-
tuations and anharmonicity prevent the excess energy from ucts of dTMP– are the neutral nucleoside and a negatively
returning in whole to the N–H bond and thus frustrate the charged phosphoric acid. This is in agreement with a number
14
Figure 18. Thymine nucleotide: the two O–H bonds on the phosphate
component are circled in red. Reprinted figure with permission from
Figure 17. The thymine nucleotide that was studied in this work [91], Copyright (2011) by the American Physical Society.
with the 3’ C–O bond within it highlighted.
of experimental studies, which have also found that DNA dam- configurations of this system. In all three cases, the bond
age can occur at these low energies because of charge transfer. extended, but quickly reformed, showing that this is not
It is interesting to note that the initial experiments of Sanche enough energy to cause dissociation. Simulations with 2 eV
et al did not observe damage in plasmid DNA by electrons broke the bonds in some cases, while 3 eV led to dissociation
with energies as low as these [1]. However, later experiments in all cases. We therefore conclude that in a micro-solvated
by the same group found DNA strand breaks were possible at environment the energy required to dissociate the bond is
energies between 0 and 4 eV, with damage peaks at 0.8 ± 0.3 much higher than the dissociation energy in gas phase, and it
eV and at 2.2 eV [42]. Simons [73] proposed an explanation may increase further in the fully solvated environment.
for this very low energy strand break in DNA in terms of a
vibronic coupling between the π ∗ orbital of the nucleobase 3.2.3. Condensed phase. The condensed phase model con-
and the σ ∗ orbital located in the C–O bond. This coupling sists of a single dTMP molecule and 64 surrounding water
allows the electron to transfer its energy to the C–O bond and molecules in a periodic simulation cell. In the gas phase,
cause bond cleavage. dTMP includes a phosphate component with the two O–H
A vibrational analysis of dTMP identified the C–O bond as bonds (Oi-H and Oj-H) that are circled in red in figure 18, i.e.
a clean normal mode, with very little weight in other parts of it is doubly protonated. During condensed phase equilibration
the molecule. Therefore, the additional energy was introduced the neutral nucleotide proved itself to be prone to deproto
by modifying the velocities of the C and O atoms using the nation at the Oi-H or Oj-H sites. The molecule therefore
method described above, while adding the extra electron to probed singly protonated configurations. To investigate the
the LUMO. In the dynamical simulations we observed that protonation state of the phosphate, we calculated the shortest
dTMP will dissociate when only 0.7 eV of additional kinetic distances between the two oxygens on the phosphate and all
energy is injected into the C–O bond. In the simulation the the hydrogens in the system to which they could form bonds,
excess electron is seen to transfer from the nucleobase to the including all the hydrogen atoms in the water molecules.
C–O bond to facilitate its cleavage. The evolution of Mulliken Figure 19 shows these distances as a function of time. Covalent
charges shows that, after cleavage, the excess electrons stays O–H bonds should be approximately 1 Å in length, so it is
with the phosphate, forming a phosphoric acid anion, which is clear from this figure that the phosphate is more often singly
consistent with the picture portrayed by the dissociation ener- protonated than doubly protonated.
gies presented above. Dissociation also happens at all higher Dynamical simulations were run on both types of configu-
energies. ration, singly and doubly protonated, following the procedure
described above. These simulations were started from con-
3.2.2. Microsolvation. The microsolvated system consists of figurations taken from the equilibration. Doubly protonated
a dTMP molecule and the 19 surrounding water molecules configurations dissociated when an excess kinetic energy of
that complete its first solvation shell [97]. An excess elec- 1 eV was introduced to the C–O bond. Most of the singly
tron was added, and simultaneously an excess kinetic energy protonated configurations required at least 2 eV to dissociate.
of 1 eV was given to the C–O bonds of three equilibrated In the singly protonated configurations that dissociated with
15
Running Average Lowest distances between Phosphate (O’s) and Hydrogens in the syste doubly protonated nucleotide we calculated the dissociation
3 energy of the singly protonated nucleotide for the following
gas phase reaction
Oj - H distance (Running Average)
Oi - H distance (Running Average) dTMP− → dThd + HPO−
(14) 4
2.5
and obtained values of 3.47 and 3.41 eV for the Oi-H Oj-H
sites, respectively. These numbers are significantly larger than
DIstance (Å)
2
the dissociation energy for doubly protonated dTMP which
was 0.08 eV. Thefore, if the nucleotide is in a singly proton-
ated state, it is unlikely to dissociate with energies below 3 eV,
which is consistent with the results from the dynamical simu-
1.5
lations that were presented above.
The literature presents some ambiguity associated with the
DEA reaction in solvated DNA. It is generally believed that
1
electron attachment to DNA bases is highly likely when DNA
0 1 2 3 4 5 6
Time (ps) is in solution as there is a large amount of evidence to support
this [91, 98–104]. However, contrasting results suggest that
Figure 19. Running averages for the shortest distances between
DNA damage is less likely to occur in solution, as dissociation
phosphate oxygens and the hydrogens atoms in the system. O–H
covalent bonds should be approximately 1 Å in length, so this barriers for the C–O bonds are higher in solution than in gas
figure suggests that the phosphate group is more likely to be singly phase DNA [72, 105–108]. The results presented here, which
protonated. Reprinted with permission from [138]. Copyright show that deprotonation of the phosphate disfavours the C–O
(2012) American Chemical Society. bond cleavage, provide an explanation for this ambiguity and
give further support for the conclusion that nucleotide disso-
1 eV, the phosphates quickly re-gained a second proton when ciation is less likely in condensed phase than in gas phase.
the energy was introduced. This protonation from the water
following dissociation seemed to prevent the C–O bond from
reforming. Looking at the spin density, we observed that in the 3.3. Larger DNA models: polynucleotides and double
doubly protonated case at 1 eV the excess electron transfers to strands
the C–O bond, which breaks. In the singly protonated system, At high electron energies, resonances are likely to lead to
by contrast, the electron remains on the nucleobase and the bond breaks and the cleavage of the 3’ phosphodiester bond
bond does not break. in particular. In DNA, however, nucleotides are part of the
It is well known that in aqueous solution nucleotides (and double helix scaffold. This ensures that they are bonded to
more generally DNA) are prone to deprotonation at the phos- other nucleotides through 5’ phosphodiester bonds as shown
phate site [98]. To make further progress the following ques- in figure 3 and that the bases are linked to the bases of the
tions must be investigated: (1) Will the phosphate component complementary nucleotide in the other strand, via hydrogen
of the nucleotide be singly or doubly protonated in condensed bonds. The consequence of all this is that DNA has quite a
phase? (2) Is the dissociation energy of the singly prot rigid structure. Furthermore, because any phosphate excised
onated nucleotide higher than that of the doubly protonated via the cleavage of the 3’ bond remains attached to its suc-
nucleotide? cessive nucleotide via a 5’ bond, it will not have an opportu-
To answer the first of these question we considered the nity to move too far away from the cleaved sugar component.
protonation state of the phosphate component from a chemi- There is, thus, a distinct possibility that the cleaved 3’ bond
cal point of view. The phosphate component of DNA is based will be re-formed after only a short period of time. A view of
on the inorganic acid H3 PO4 (phosphoric acid), which can a solvated fragment of DNA made of 12 base pairs is shown
undergo the three deprotation reactions shown below: in figure 20.
H3 PO4 H+ + H2 PO− ; pKa = 2.12 What keeps single-stranded polynucleotides intact are the
4
(non-bonded) stacking interactions between the bases. These
H2 PO− 4 H + +
HPO2−
4 ; pKa = 7.21 are weak van der Waals forces, but they may still be suffi-
H + HPO2−4 ; +
PO3−
pKa = 12.67
4 cient to maintain the the structure of the polynucleotide even
(13) in the presence of a strand break. These interactions may
In the nucleotide, one of the three protons of phosphoric acid therefore favour the reforming of the bond. Obviously, these
is replaced by a carbon atom which joins the phosphate to interactions will not help single-stranded DNA to repair itself.
the rest of the nucleotide. The pKa value for the second of the Single stranded DNA forms occur during replication, when
reactions above suggests that release of a second proton from the H-bonded network between the two strands is unzipped by
the doubly protonated nucleotide is also likely. The release of the action of a combination of enzymes, and a second strand
the third proton is much less likely so the fact that we observe is synthesised by DNA polymerase. It may therefore be more
the nucleotide in a single protonated form seems reasonable. difficult to repair strand breaks if they occur during the rep-
To answer the second question about whether the energy lication phase, which will happen more often in frequently
of the single protonated nucleotide is higher than that of the dividing cells.
16
for the capture cross section. Gas-phase calculations indicate

that the pyrimidines thymine and cytosine are favoured. In the
condensed phase, however, environmental fluctuations can
modify this behaviour, as shown in figure 22.
The study of DEA in a double-stranded situation requires
the identification of which bonds are coupled to the reso-
nant electron, as a function of energy, which is a highly
non-trivial matter. Once these bonds have been identified,
one could attempt a simulation approach similar to that used
for the nucleobase or the nucleotide, above. In such cases,
a single strand break will alter the spatial configuration of
the helix, twisting it in unnatural ways that can be detected
by the enzymatic repair machinery. The question here is
whether the phosphodiester bond will reform naturally, or
whether there is some aspect in the way it breaks, that pre-
vents it from reforming. If the geometric distortion of the
helix is too large, then it is unlikely that the molecule will
be able to find the way back to the repaired state. On top of
this the strand break may result in chemical modifications to
Figure 20. A fragment of DNA composed of 12 base pairs,
and solvated in water. This is the smallest stable fragment, and the phosphate and the sugar that raise the free energy barrier
comprises two turns of the helix. Reprinted with permission from to bond reformation. These hypotheses and their likelihood
[139]. Copyright (2015) American Chemical Society. can be assessed using simulations similar to those described
in section 3.2.
In single-stranded DNA (or RNA) fragments, a particularly
interesting question is whether there are specific nucleobase
sequences that favour the cleavage of the glycosidic and phos- 4. From DEA to thermal activation
phodiester bonds. This matter was addressed experimentally
by Li et al [109] in TXT trinucleotides, where T is thymine In [73], Simons described a mechanism to explain how very
and X is any of the four bases. Both types of bonds were low energy electrons cause strand breaks in DNA. In Simons
observed to cleave, and trends were identified according to the mechanism low energy electrons attach to the π ∗ orbital of the
identity of the central base X. Simons’ argument of a two step nucleobase and form a shape resonance. They then transfer to
process for electronic transfer and bond breakage [73] can a σ ∗ orbital of the sugar-phosphate C–O bond. This transfer
be extended to the case of polynucleotides. Excess electrons relies on thermal vibrational motion of the neutral molecule
attach at the base, and when a phosphodiester bond stretches and also on the lifetime of the resonance. During the ther-
it transfers and cleaves the bond. For a trinucleotide there are mal motion, the C–O bond vibrates around its equilibrium
four possible such bonds that can cleave, apart from the glyco- bond length of 1.45 Å. If an electron with energy ∼1 eV is
sidic ones. These are shown schematically in the left panel of resonantly attached to the molecule when the C–O bond is
figure 21. The other two panels illustrate the previous concept. vibrating but at a length smaller than 1.8 Å, the electron will
Before cleavage the excess electron, whose probability distri- be captured in a base π ∗ orbital. If, within the lifetime of the
bution is represented here by a blue surface, is located at one resonance, i.e. before the electron detaches, the C–O bond
of the end bases, in this case thymine (middle panel). When stretches beyond 1.85 Å, then the σ ∗ state becomes lower in
the C–O bond is stretched, the excess electron transfers to the energy than the π ∗ orbital, the electron migrates towards the
phosphate involved in this bond (right panel) [110]. σ ∗ orbital, and breaks the C–O bond. The C–N glycosidic
Related experimental work was carried out earlier for GCAT bond is also susceptible to break, but the barrier is higher.
tetranucleotides in which G or A had been removed [111]. In The likelihood that DNA damage would occur at one of
this work it was shown that the capture probability depends on the C–O bonds at such low electron energies was confirmed
the sequence, while bond cleavage is not significantly affected by experiments performed in Sanche’s group. These authors
by the base deletion [112]. Experimentally it is very difficult exposed oligonucleotides (CGTA and GCTA) to near-zero eV
to disentangle electron capture from bond cleavage at the level electrons and discovered that cleavage was primarily along
of cross sections, as gel electrophoresis distinguishes only the the C–O backbone [113]. These findings led to numerous
final products. From the theoretical point of view, under DEA studies of strand breaking reactions in ground state situations,
conditions the electronic and vibrational states are correlated, under conditions where bond cleavage is reliant on thermal
and so are the capture and cleavage cross sections. If, how- vibrational motion. These studies focus on the ground state
ever, caging effects due to the environment prevent the direct electronic structure of DNA anions whilst activating the
cleavage, then capture and bond cleavage become independent bond stretching. This is generally done by fixing the value of
events. This will be discussed in next section. a relevant reaction coordinate, generally a bond length, and
In double-stranded DNA the question arises of where is the computing the energy as a function of the reaction coordinate
excess electron going to localize in a base pair. This is relevant via constrained optimisation techniques. This allows for the
17
Figure 21. TGT trinucleotide. Left panel: schematic representation showing the four possible phosphodiester bonds that can be cleaved
(numbered 1–4). Middle and right panels show the location of the excess electron (blue surface) before and after the cleavage, respectively.
Reprinted with permission from [139]. Copyright (2015) American Chemical Society.
Figure 22. Left panel: AT (top) and GC (bottom) base pairs in the gas phase. The blue surface represents the probability distribution of
the excess electron. Right and central panels: same as left, but in a solvated environment. Delocalization between the two bases can be
observed. Reprinted with permission from [139]. Copyright (2015) American Chemical Society.
computation of activation barriers and detachment energies then a thermodynamic equilibrium will be established after a
at zero temperature, without considering the thermal fluctua- short period. In that case the system becomes a ground state
tions of the environment. Numerous theoretical studies using anion immersed in a fluctuating environment. Under these cir-
both wavefunction and DFT methods support the notion that cumstances, DNA strand breaks can still occur, but as ther-
strand breaks can be triggered by such low energy electrons. mally activated processes, controlled by a free energy barrier.
Bao et al [72, 105] calculated the activation energies for bond The thermal activation mechanism is as follows: as the
cleavage in the pyrimidine nucleotides and found that C–O environment experiences thermal fluctuations at the temper
strand breaks should dominate at near-zero energies. By ature of interest (room temperature in our case), the bond in
calculating the detachment energies from the base they also question randomly performs excursions away from its equi-
showed that it is feasible that electron attachment to the nucle- librium position. When the bond is stretched, the energy will
obase can trigger bond cleavage [114]. Barrios et al explored increase and the bond will be put in an unfavourable situation.
the energy surfaces leading to bond cleavage and also found The Boltzmann factor ensures, however, that if that energy
that C–O cleavage would be operational at such low energies increase is not too high, these fluctuations in bond length
[41, 115]. carry a non-negligible probability. The larger the stretch, the
The results in sections 1.4 and 3 highlight the importance of higher the energy and the lower the probability. Occasionally,
including the aqueous environment when considering DEA in the bond can stretch to the point in which the energy reaches
DNA. The environment stabilizes the resonance increasing its its maximum, i.e. the transition state. Beyond that point, the
lifetime, thus buying time for energy transfer to occur, which bond breaks and the system evolves towards a product state
is consistent with the mechanism proposed by Barrios et al that consists of two fragments. In the present case, one frag-
[41]. In addition, the energy required to break a bond is higher ment will be negatively charged, and the other will remain
in solution than in gas phase because of caging effects. If the neutral. Furthermore, one will be a radical, and the other a
situation is such that the excess energy of the resonant electron closed-shell molecule.
is transferred to the C–O bond and if the environment does not The description of this process is different from that of
allow the bond to break because it reflects vibrational energy DEA. We have to start from the attachment of the excess
back into the nucleotide and dissipates it into the environment, electron to the DNA fragment, i.e. the insertion of an electron
18
Table 1. Adiabatic electron affinity of uracil: ab initio. Table 2. Adiabatic electron affinity of uracil: DFT.
Theory level AEA (eV) Functional AEA (eV)

HF −1.00 [119] GGA
MP2 −0.21 [120] BLYP 0.15a [122]
CCSD(T) −0.05 [120] PBELYP −0.18 [123]
CASPT2 0.03 [120] Hybrid
Experimental method AEA (eV) B3LYP 0.24a [122]
RET 0.06 ± 0.03 [118] M05-2X 0.17 [36]
PS 0.15–0.16 [117] M06-2X 0.11 [36]
Experimental reference
into the LUMO of the neutral fragment and the conversion PS 0.15 ± 0.12 [117]
of this orbital into a singly-occupied molecular orbital (or a
Includes zero-point vibrational energy correction.
SOMO). A first question is where is the SOMO spatially
located, and whether it is similar or different to the LUMO.
Secondly, we must ask whether the neutral geometry is suit- Theoretical studies of electron attachment to DNA subunits
able for binding an additional electron or whether a struc- were pioneered by Colson, Belson, and Sevilla approximately
tural reorganisation is required. Then, we need to identify 25 years ago using the Hartree–Fock (HF) approximation
the bond (or bonds) that are prone to break. That is to say [119]. Subsequently, EA values were refined using post-HF
we need to identify the bonds that are weakened by the addi- methods that introduce electron correlation at different levels,
tion of the electron so that we can study the energy (at zero including second order Møller–Plesset perturbation theory—
temperature) or free energy (at finite temperature) profile as MP2—, coupled-clusters including perturbative triple exci-
a function of the length of these bonds. In this section we tations—CCSD(T)—, and multiconfigurational perturbation
will focus on these questions. theory—CASPT2. The AEA values of uracil calculated using
ab initio methods are presented in table 1 and are compared
to the experimental values. It should be pointed out that the
4.1. Electron affinities theoretical values include a zero-point vibrational correction.
The electron attachment properties of DNA subunits can be It is clear that HF severely underestimates the AEA of uracil.
investigated through the electron affinity (EA). This measur- MP2 and CCSD get closer to experimental values with large
able quantity is the difference between the energy of a neu- basis sets. The most accurate molecular orbital approach,
tral molecule and the energy of the anion following electron however, is CASPT2.
attachment (TNI). It is important to remember that these methods are very
sensitive to the choice and the size of the basis set that is
EA = Eneutral − Eanion .
(15) used to represent the molecular orbitals, especially for calcu-
The more positive the EA is, the more likely the molecule will lating EA. Anions are frequently loosely bound and require
be to attach an electron. The anion and the neutral molecule diffuse basis orbitals and large basis sets. In fact, the addi-
have two different potential energy surfaces. The characteris- tional electron is often not bound so its wave function should
tic attachment energies, i.e. the difference between neutral and be a plane wave. If the basis set used in simulations of such
anion energies, depend on the experiment performed, but are systems is composed of local orbitals like Gaussians, it will
usually one of the following three quantities: produce an artificially bound state at a positive energy that
will become increasingly closer to zero as the size of the basis
(i) Vertical attachment energy (VAE): the difference in the
set is increased. This is a signature that something is wrong
energies of the anion and the neutral molecule for the
with the calculation. It happens mostly with the VAE, as most
optimised geometry of the neutral molecule.
molecules struggle to bind an additional electron at the neu-
(ii) Vertical detachment energy (VDE): the difference in the
tral geometry. It is less relevant for the AEA and the VDE, as
energies of the anion and the neutral molecule for the
the anion is already at its optimised geometry, and the neutral
optimised geometry of the radical anion.
molecule is usually stable in its own geometry and in that of
(iii) Adiabatic electron affinity (AEA): the difference between
the anion.
the energy of the optimised anion and the optimized neu-
Whilst advanced wavefunction methods are highly accu-
tral molecule.
rate, their application is limited to relatively small system
Experimentally, the two most commonly used methods sizes. Density functional theory (DFT) is an attractive alterna-
to measure AEA are photoelectron spectroscopy (PS) and tive because of its reliability and affordability. This method
Rydberg electron transfer (RET) methods [116]. Experimental allows one to model relatively large molecular systems with
AEA values provide benchmarks for theoretical approaches. a reasonably high accuracy. DFT calculates the ground state
According to PS, the AEA of uracil is 0.15–0.16 eV [117], energy of a system as a functional of its electronic density while
while the RET value is 0.06 ± 0.03 eV [118]. Positive values approximating the exchange and correlation (XC) component
suggest that electron attachment is favourable. of the energy functional. The simplest XC approximation is
19
Table 3. Adiabatic electron affinities (eV) for the four nucleobases at different levels of theory (all values include the zero-point vibrational
contribution).
Ab initio
Method Thymine Cytosine Adenine Guanine
MP2 [120] −0.26 −0.4 −0.71 −1.06
CCSD [120] −0.09 −0.17 −0.84 −0.44
CASPT2 [120] 0.02 −0.10 −0.44 −0.72
Density functional theory
M05-2X [36] 0.12 −0.05 −0.35 −0.54
M06-2X [36] 0.07 −0.08 −0.36 −0.54
B3LYP [122] 0.20 0.03 −0.28 −0.10
BLYP [122] 0.12 −0.01 −0.19 −0.01
PBE0 [128] 0.133 0.02 −0.024 −0.04
Experiment
PS 0.15 ± 0.12 [117] 0.10 ± 0.12 [117] 1.51 ± 0.05 [129] 0.95 ± 0.05 [129]
the time-honoured local density approximation (LDA), which and closer to zero. However, whilst having slightly positive
produces significant errors when computing AEAs [121]. The predicted AEA values, gas phase cytosine is an unlikely host
next rung in the ladder is the generalised gradient approx for the excess electron and as such thymine is the most likely
imation (GGA), that includes the density and its gradient. electron host as it has the highest AEA values.
AEA for DNA and RNA bases have been computed using the In summary, the pyrimidines, i.e. uracil, thymine, and
BLYP and PBELYP functionals. Values for uracil are reported Cytosine, are the bases with the largest electron affinities
in table 2. The BLYP affinity is fortuitously close to the PS and are thus the most likely to attach an electron. Guanine
experimental value. and Adenine exhibit negative values and are therefore
More accurate functionals are constructed by combining unlikely to capture an electron. If we put the bases in order
GGAs with HF (exact) exchange in an appropriate propor- of decreasing AEA, we obtain the following sequence:
tion. This is the case for the hybrid B3LYP functional [124, U T > C > A ∼ G.
125], which has proved itself to be particularly reliable for
predicting AEAs of DNA components [122]. A further step
4.3. Solvation effects
up is to combine the hybrid exchange with a meta-GGA that
includes not only the gradient of the density, but also the The low-energy interaction and dissociation mechanisms
kinetic energy density (laplacian). This formalism is used in which have been described in the previous sections were
the functionals that belong to the Minnesota suite (M05-2X mostly investigated by performing experiments and calcul
and M06-2X [126, 127]). These have recently been shown ations on dry (gas-phase) DNA fragments. However, in a
to provide the best comparison to the references in electron living cell DNA is in a condensed phase environment and is
attachment calculations. AEAs for uracil computed with the surrounded by other components including water which is the
various functionals are reported in table 2 and are compared most abundant species in the cell and which plays a vital role
against experimental reference data. in maintaining DNA structure and function [130, 131]. In both
theory and experiment, LEE interactions with DNA are modi-
fied in the presence of the aqueous environment. During all
4.2. Base specificity
the stages of a chemical reaction, the solvent interacts with the
Uracil is the nucleic acid with the smallest number of atoms solute and thus influences the thermodynamics and kinetics of
so it is widely used in preliminary calculations to make the reaction. In computer simulations, the effect of solvation
comparisons between theoretical methods so that a suitable can be introduced either explicitly or by using a continuum
method for examining all other nucleobases and larger units model. Discrete (or explicit) solvation involves direct addition
can be settled on. Interestingly, there are significant differ- of water molecules to the solute model. Polarizable contin-
ences in the electron affinities of the four DNA bases which uum solvation models (PCM), sometimes also called reaction
are consistent across all levels of theory. Table 3 summarizes fields, cut the computational cost of considering each molecule
the AEA values that have been obtained for all the bases in the solvent explicitly by placing the molecule in a cavity
using the various methods described in section 4.1. Whilst the carved in a medium characterised by a dielectric constant. In
actual values of the AEAs of each of the bases differ with each the simplest PCM, the dipole moment of the molecule, which
method, the trend in the ordering is consistent. The purine is placed in a spherical cavity, polarises the medium, which
bases (Gua and Ade) have the lowest electron affinities. The in turn produces an electric field in the cavity and leads to an
extensive theoretical review by Gu et al states that ‘electron effective electrostatic interaction between solute and solvent.
attachment to gas-phase guanine is not viable and that ade- More sophisticated models that take into account the shape of
nine in the gas phase would not be a good electron acceptor’. the molecule and higher moments of the charge distribution
[36]. The electron affinity of cytosine is predicted to be higher [132] are nowadays implemented in most quantum chemistry
20
post-HF, coupled cluster (CCSD) methods, that the AEA of

uracil would increase upon addition of up to three water mol-
ecules. They calculated a predicted value of 0.42 eV for the
tri-hydrated uracil cluster [103]. The effect of an increased
number of water molecules in the solvation shell, up to 15,
was computed at the DFT-PBE0 level of theory [91]. In line
with Kim et al, the AEA of each of the bases increases with
the number of molecules and saturates at values of around
1 eV, as shown in figure 23.
Schiedt et al [117] investigated electron attachment to
micro-solvated uracil, thymine and cytosine using photode-
tachment-photoelectron (PD-PE) spectroscopy experiments.
The electron affinities obtained from the PD-PE spectra
showed that the average increase per added water molecule
was 0.22 eV for uracil, 0.20 eV for thymine and 0.17 eV for
cytosine. These experimental values are consistent with the
calculated ones.
4.3.2. Solvation effects on dissociation. The electron capture

ability of DNA fragments increases as the level of solvation
increases but there is ambiguity surrounding the dissociation
of these molecules in solution. In some experiments, hydration
can be introduced to films of DNA as monolayers. Ptasinska
and Sanche [133] performed anion desorption experiments on
short DNA single strands covered by three water monolayers.
The anion yield, or in other words the dissociation products,
were increased by a factor of 1.6 when in this solution. Fur-
Figure 23. AEA values for each of the nucleobases with an
increasing number of surrounding water molecules. Reprinted with thermore, in recent experiments by Alizadeh et al [28], the
permission from Smyth and Kohanoff [91]. Copyright 2011 by the hydration level was varied up to a bulk-like water environ
American Physical Society. ment. The amount of DNA damage increased significantly in
bulk-like water because of the increase in the AEAs of DNA
codes. PCMs proved effective in studying electron attachment bases when in solution. However, ambiguity remains because
to DNA, but they do not include specific interactions such numerous studies that investigated the influence of a water
as hydrogen bonds between solute and solvent molecules. environment on dissociation barriers suggest that dissociation
H-bonding is known to stabilize the excess electron in the should be more difficult in solution.
solute, thus increasing the electron affinity [91]. To overcome Studies by Gu et al [72, 105–107] at the DFT-B3LYP
this issue, PCMs can also be combined with explicit solvation level of theory investigated solvation effects on the cleavage
methods by considering a first explicit solvation shell embed- of the pyrimidine strands (C and T) by using a polarisable
ded in a polarisable continuum. continuum to model solvation. These studies demonstrated
that the activation barriers for C–O and C–N bond cleavage
4.3.1. Solvation effects on electron attachment. AEA values increase significantly when the molecule is moved from a gas
for DNA bases are known to increase in the aqueous environ phase to a solvated environment. In other words, the aque-
ment so solvated DNA is a better electron host. Gu et al [98], ous environment greatly increases the activation energies for
employed DFT-B3LYP calculations to investigate the influ- strand breaks at the C3-O3, C5-O5, and for base excision at
ence the aqueous environment has on electron attachment to the glycosidic bond. Interestingly, guanine displays a different
single stranded DNA. In these simulations the electron cap- behaviour to the other nucleobases. Schyman and Laaksonen
ture ability increased significantly when solvent effects were [134], used DFT-B3LYP to investigate the effects of both the
included via a PCM. In gas phase, the AEA values of the DNA PCM and an explicit water environment on C3-O3 cleavage
nucleotides were within the range 0.1–0.35 eV whereas, in barriers in guanine-3’-monophosphate (dGMPT). They found
aqueous solution, the AEA values increased to 0.95–1.99 eV. that cleavage barriers were lower when in solution compared
The influence of discrete microsolvation on the electron to those calculated in the gas-phase. Further studies by Gu
capture ability of the DNA base thymine has been extensively et al of the guanine-3’-5’-monophosphate found activation
investigated by Kim et al [101], who performed DFT-B3LYP barriers for the cleavage of C3-O3, C5-O5 and C–N bonds
calculations with up to 5 water molecules. They found that were dramatically decreased upon solvation by a PCM [135].
AEA values increased gradually with the number of hydration These findings help explain the experimental observation
molecules and hence that the hydrated anions were more sta- that guanine might be the weak link in DNA. Time-resolved
ble than the gas phase anions. AEA values increased to 0.91 eV laser spectroscopy (TRLS) was used by Wang et al [136] to
with 5 water molecules. Dedikova et al also predicted, using investigate low-energy electron interactions with solvated
21
nucleotides. They found that electron attachment to all nucle- as hydrogen bonding. What they do not generally do, unless
otides would occur within 1 ps. Furthermore, these authors reactive force fields such as Morse potentials are used, is allow
found that both dAMP and dCMP nucleotides formed stable for water dissociation and proton transfer. In order to over-
anions after electron attachment. However, 35% of the dTMP come these limitations, free-energy barriers were computed
anion and 60% of the dGMP anion dissociated. In other in a fully quantum model for each of the four DNA nucle-
experiments, Zheng et al investigated strand breaks in short otides surrounded by 100 explicit water molecules [138].
DNA sequences (GCAT), where individual bases were deleted In this work, the gas-phase barriers for the cleavage of the
selectively to understand their roles. Again, guanine was high- C3 -O3 bond were compared to those in static solvation and
lighted as a particularly sensitive damage site because the in a thermally fluctuating solvated environment. The static
extent of damage decreased significantly when G or A were solvation barriers were the highest by a significant margin,
removed [111]. These results will be discussed again later whereas gas-phase and thermally fluctuating barriers were
within the context of free energy calculations. comparable. It subsequently emerged that in that paper there
The effect of the direct interaction between water and DNA were some problems with the statistical analysis so the free
components was also highlighted in DFT studies by Kumar energies for the thermally fluctuating systems were underes-
and Sevilla [94], who modelled the cleavage of 5’-thymine timated approximately by a factor of two. This was rectified
monophosphate with and without explicit water molecules. by McAllister et al [139], who reported barriers between 15
Their results showed that the cleavage barriers increased sig- and 20 kcal mol−1 for all nucleotides. This suggests that the
nificantly upon solvation. Furthermore, there was a suggestion energy barriers calculated by optimising the geometry for dif-
that some of the waters were hydrogen-bonded to the nucleo- ferent values of the reaction coordinate (static barriers) can
tide in their optimised structures. be significantly larger than those calculated dynamically, thus
highlighting the importance of bringing environmental fluc-
tuations into the picture [138].
4.4. Molecular dynamics simulations and free energies
In addition to the free energy barriers, these simulations
While static (zero-temperature) calculations are useful to provided a clear proof that the strand break mechanism pro-
establish the foundations for understanding DNA damage in posed by Simons [41, 73] still operates in the condensed
solution, they do not take thermal and electrostatic fluctuations phase. Evidence for this assertion is provided in the spin den-
in the environment into account, which are non-negligible at sity plots shown in figure 24. These images clearly demon-
room temperature. Moreover, activation energies are generally strate that the excess electron is captured at the base. Then
computed along a specific, one-dimensional reaction coordi- when the phosphodiester bond elongates to about 1.8 Å, the
nate, such as the length of a C–O bond. Molecular Dynamics electron delocalizes over the whole nucleotide. This electronic
(MD) simulations provide a way of incorporating the effect of state is a combination of the two crossing states, that become
these fluctuations. The water component can be treated at the degenerate at the transition state. For larger elongations, the
same theory level of the DNA fragment, or otherwise using excess electron re-localizes on the sugar-phosphate region and
a classical force field within a widely used methodology that the free energy starts decreasing towards the products basin.
has been called QM/MM. First-principles MD simulations Upon further elongation, the bond eventually breaks.
(FPMD) where water and solute are all treated at the DFT-
PBE level, have been used to simulate the dynamics of the 4.4.1. Protonation. The authors of [138] noticed that adenos-
excess electron in an explicitly solvated nucleobase system ine monophosphate (dAMPT) behaved differently from the
with 64 water molecules [91]. In pure water an excess elec- other nucleotides, and required twice as much energy to sur-
tron will roughly localise in a cavity with the water hydrogens mount the cleavage barrier. Upon closer scrutiny, it became
pointing towards it [137]. However, when a solute is present clear that the reason for this enhanced barrier height was that
the electron localises very rapidly on the nucleobase, over a the base spontaneously protonated at one of the N sites, by
timescale on the order of 15–25 fs. This fast localisation on seizing a proton from a nearby water molecule. Notice that
the nucleobase was also observed for nucleosides and nucleo- these kind of events are not possible if the water molecules
tides [128, 138]. are represented by a continuum medium, or with explicit clas-
Free energy barriers for the phosphodiester bond in 3’ sical force fields that do not allow for dissociation of the O–H
and 5’ cytidine monophosphate (dCMP) were computed bonds. It is not obvious either that this can be observed in
by Schyman et al [108], using a QM/MM model in which a hybrid explicit-quantum/PCM model, as the protonation
the nucleotide was described at the DFT-BLYP level and water may not necessarily be one of those treated explicitly.
immersed in a classical (AMBER) water bath containing over The explanation given for the increase in the height of the
1000 water molecules. The authors of this study computed barrier was that the presence of the excess electron on the
barriers for phosphodiester bond cleavage on the order of base attracted the water molecule and thus favoured proton-
30 kcal mol−1, which are much higher than the gas-phase ation. Once the base was protonated, this very same proton
barriers. Explicit classical water molecules provide a much stabilised the excess electron at the base. Since the cleavage
better representation of the environment than a polaris- is related to the crossing of two electronic states, one located
able continuum, as they include specific interactions such on the base and another on the sugar-phosphate [73], the
22
Figure 24. Spin density plots for dTMPH−: (left) at the equilibrium bond length of 1.4 Å; (middle) at the transition state, with a bond
length of 1.8 Å; (right) after bond cleavage, at a bond length of 2.2 Å. Reprinted with permission from Smyth and Kohanoff [138].
Copyright 2012 American Chemical Society.
study suggested that protonation stabilised the anion and that,

if this was the case, C–O bond cleavage might be less likely
than cleavage of other bonds in the system. A consideration
of H-bonding and protonation was thus deemed crucial for
determining reaction pathways and barriers.
The likelihood for protonation of nucleotides in aqueous
solution, and the effect of protonation on free energy barri-
ers, were thoroughly studied for all four nucleotides [139,
140]. This was achieved by means of FPMD simulations in
which the lengths of specific bonds were constrained to a
series of increasing values. Thermodynamic integration of the
constraint force was then used to obtain the free energy pro-
file (blue moon ensemble). Each nucleotide was placed in a
15 Å periodic box with 100 explicit water molecules, with the
electronic structure calculated at the DFT-PBE level.
To study protonation a reaction coordinate that measured
the shortest distance between an N or O in the nucleobase,
and a hydrogen in a water molecule was used. It is impor-
tant that the hydrogen in this coordinate can belong any water
molecule in the system and that a hydrogen belonging to the
water molecule that instantaneously is closest to the substrate
is not used as the identity of the closest water molecules to the
base can change during the simulations [140]. The systems
studied are depicted in figure 25 and the free energy profiles
Figure 25. The four nucleotides studied in [139], showing the as a function of the distance between the protonation sites and
closest water molecule to each base (circled in red). Protonation the hydrogen atom belonging to the hydrogen-bonded water
was enforced for those hydrogen bonds. In the case of dAMPH−,
the water molecule circled in red spontaneously protonated the
molecule are shown in figure 26.
nucleobase during equilibration and the shortest hydrogen bond was Figure 26 shows that protonation is favourable for
formed by the water molecule dash-circled in blue. Reprinted with dCMPH– and dTMPH–, and barrier-less for dAMPH–. This
permission from McAllister et al [139]. Copyright 2015 American helps to explain why adenine spontaneously protonated dur-
Chemical Society. ing unconstrained equilibration in [138], and confirms that it
should be in this protonated configuration when in the con-
stabilisation of the electron in the π ∗ orbital due to proton- densed phase. There are two minima in the free energy pro-
ation shifts the crossing of π ∗ and σ ∗ energies to longer C–O files for protonation of cytosine and thymine, one where the
distances—distances on the order of 2 Å rather than 1.8 Å. nucleobase is protonated and one where the hydrogen atom is
In those additional 0.2 Å, the free energy continues to rise covalently bonded to a water molecule and hydrogen bonded
reaching double the value at the transition state (the maximum to the nucleobase. These minima are separated by a very small
of the free energy curve). These results are consistent with a barrier, under 1 kcal mol−1. Protonation of dGMPH– is unfa-
study of the protonated cytosine structure by Gu et al [98], in vourable, and the minimum in the free energy occurs when the
which protonation was enforced at the N3 nucleobase site and nucleobase is hydrogen-bonded to its closest water molecule.
an analysis of its behaviour in a base pair was performed. This Hence, dGMPH– will not protonate in the condensed phase.
23
There is a considerable amount of evidence suggesting that

strand breaks are most likely to occur at the guanine site on
the DNA strand [106, 111, 134–136]. In particular, Wang et al
found that dissociation by the addition of electrons to the gua-
nine base, results in significant numbers of single and double
strand breaks [136], which is in agreement with the above
results. Figure 26 shows that guanine will be deprotonated,
whereas figure 27 shows that the barrier for breaking the
C3-O3 bond in deprotonated dGMPH– is only 19 kcal mol−1
and that the total free energy change upon reaction is approxi-
mately zero. The barrier for protonated dAMPH– is simi-
lar to that of dGMPH–, and significantly lower than those
of dCMPH– and dTMPH–. In other words, strand breaks
appear to be more likely at the purine than the pyrimidine
sites, which is consistent with the results of Zheng et al who
Figure 26. The free energy profiles for the four nucleotides studied observed that the extent of damage to short polynucleotide
in [139]. The vertical dashed line indicates the typical length of sequences (GCAT) was lower when guanine and adenine were
a bond involving a hydrogen atom and, thus, indicates the point absent.
at which the nucleobase becomes protonated. It is clear that with
the exception of dGMPH−, the free energy of the nucleotides is
4.4.2. Double strands. A QM/MM study of the molecular
minimized when it is protonated. Reprinted with permission from
McAllister et al [139]. Copyright 2015 American Chemical Society. reorganisation of solvated DNA double strands was recently
conducted by Cauet et al [142]. Calculations were performed
These calculations suggest that upon electron attachment, on 12-mer DNA fragments where two nucleotides were
in the condensed phase, many of the bases will become proto modelled at the DFT-M06-2X level of theory (hybrid DFT-HF
nated. To elucidate the effect of protonation on the strand functional), and the rest of the molecule, the environment and
breaking reactions, free energy profiles were computed by the counterions were treated at a classical (AMBER) level.
constraining the bond length between the protonation sites These authors found that attachment would occur on the sugar-
and hydrogen atoms used in the protonation calculations (see phosphate backbone, and that the P–O and C–C bonds were
figure 25). This was done for two situations, protonated and more susceptible to cleavage than the C–O bonds. However,
non-protonated nucleotides. For the protonated nucleotide the they did not rule out a transfer of energy to the C–O bond.
bond length was fixed at 1.05 Å, and for the non-protonated Apart form the possible limitations of the QM/MM approach,
case, to 1.8 Å. The non-protonated nucleotide was constrained a main difference between these simulations and those
to ensure that none of the nucleotides would spontaneously described above is that these calculations were performed for
protonate during the strand breaks, thus skewing the compara- double-stranded DNA, while the previous work focused on
tive results. The free energy curves are shown in figure 27. single nucleotides. In double-stranded situations, the base is
Figure 27 shows that the protonation of the base gener- already hydrogen-bonded to the complementary base in the
ally makes the strand break reaction more unfavorable. In other strand. Therefore, to protonate it has to seize a proton
fact the only exception to this general rule is dAMPH–. For from the other base, instead of from a water molecule. A sug-
dCMPH– the effect protonation has is particularly significant. gestion of possible transfer to the C–O bond may provide an
For dAMPH–—a nucleotide that protonates spontaneously explanation for the confusion surrounding strand breaks in
in unbiased MD simulations—the barrier to strand breaking solution. In some experiments SSB yields are significantly
is 4 kcal mol−1 lower when the system is protonated, but is increased by the inclusion of a water environment [136, 143],
similar to the barriers for the deprotonated versions of the so it seems confusing that SSB barriers can be so much higher
other bases. This is a consequence of the fact that there is no than in gas phase. The barriers being calculated, however, are
truly stable hydrogen-bonded configuration for this system. reliant on predictions made in the gas phase. Cauet et al sug-
By enforcing a second protonation the strand breaking barrier gested that other bonds can perhaps break and contribute to
increases as in the other cases. It is possible to connect these SSB yields, especially in longer DNA fragments in solution.
results with experimental observations. For example, Wang This idea remains poorly explored, but it is certainly an area
et al showed that strand breaking is significantly less likely to for further study.
occur at the Cytosine sites [136] on a DNA strand. This result
cannot be explained by examining the free energy barriers for 4.4.3. Amino acids. Another major component of the physi-
the deprotonated systems as all barriers are similar. However, ological environment are the amino acids. DNA is wrapped
for protonated dCMPH– the barrier to strand breaks is sig- around histones (proteins) in chromatin and the close proxim-
nificantly higher than for the other bases. This, together with ity of amino acids can influence the interaction between excess
the observation that it is easy to protonate this base (see fig- electrons and DNA in various ways. There have been both
ure 26) and the experimental result that the cytosine radical experimental and theoretical investigations focusing on the
anion is extremely basic (pKa = 13) [141], explains Wang’s interactions of amino acids with DNA components, including
results. hydrogen bonding, ionic and van der Waals interactions, and
24
Figure 27. Free energy profiles for the cleavage of the C3 -O3 bond in each of the four nucleotide anions. The red lines are for protonated
nucleotides and the black dashed lines are for non-protonated nucleotides. It is clear that the protonation state of the nucleotide has an
effect on the free energy barrier. In all cases (except for adenine), the barrier to break the bond is higher when the nucleotide is protonated.
Reprinted with permission from McAllister et al [139]. Copyright 2015 American Chemical Society.
stacking and T-shaped interactions. Gas-phase studies have an excess electron was added, it initially finds itself in a pre-
highlighted that a barrier-free proton transfer occurs between solvated state, that is delocalised over the the thymine and the
anionic nucleobases and amino acids [144]. Hydrogen bonds, glycine. What happens subsequently depends on the bonding
in turn, have been shown to stabilise anionic nucleobase struc- pattern between thymine and glycine. A first possibility, which
tures in an aqueous environment, which highlighted proton is shown in the right panel in figure 28, occurs when the local
acceptor sites as particularly able to stabilise a structure [91]. environment of the thymine contains hydrogen-bonded glycine
Recently, Solomun and Skalický reported that the single molecules that are ready to transfer a proton to the thymine.
strand DNA binding Escherichia coli protein can effectively In this case, the excess electron is chemically stabilised in the
inhibit single strand breaks (SSB) of oligonucleotides induced nucleobase in a manner that is similar to the case observed for
by 3 eV LEEs [145]. Furthermore, Ptasińska et al reported the protonation from water discussed in section 4.4.1. A second
fragmentation of short DNA strands irradiated by a 1 eV elec- possibility is that glycine, which has an electron affinity that
tron beam in the presence of glycine and arginine [146]. Their is similar to that of the nucleobases, physically scavenges the
results showed that DNA damage is promoted at low amino excess electron and thus prevents it from causing damage to
acid concentrations and inhibited at high concentrations. It the DNA. This is shown in the left panel in figure 28. Both of
has been suggested that there are two underlying mechanisms these mechanisms are protective of DNA and this study repre-
influencing these findings. Firstly the protein may create sents a first step towards including the most important comp
a physical shielding to the DNA. In addition the hydrogen- onents that surround DNA in the physiological medium and in
bonding interactions may play an essential role in quenching understanding how natural selection has optimized the cellular
SSB through the stabilization of the anionic species. These medium so as to prevent DNA damage events. Further work in
hypotheses were successively tested via FPMD simulations this direction is clearly needed.
of a single thymine in a bath of 32 glycine molecules instead
of water [147]. These were fully quantum simulations at the
DFT-PBE level and TZVP-MOLOPT basis set. 5. Concluding remarks and future research
In these simulations, after a short period dominated by avenues
proton transfer, a dynamical equilibrium was established
between the canonical and zwitterionic versions of glycine and There are two main ways by which radiation can cause dam-
protonated and deprotonated versions of this molecule. When age to DNA: directly, by ionising DNA, and indirectly, by
25
These processes bounce energy between the solute and the

medium and thus serve to dissipate the vibrational energy that
the DEA event transferred into the bond.
We thus assume that when the LEE is captured at high
energy, the amount of energy that is transferred to the bond
will be sufficiently large to break it. Furthermore, this large
input of energy will allow the dissociated fragment to travel
a great distance and this substantial motion will prevent the
bond from reforming. When, by contrast, the LEE is captured
at low energy, reflection and dissipation of energy will either
prevent the bond from breaking at all or will prevent the DNA
Figure 28. Snapshot of the spin density on the thymine-glycine fragments that are formed when the bond breaks from moving
complex with an excess electron after equilibration. Left panel: apart. Consequently, when the LEE is captured at low energy
the electron is scavenged by one of the glycines. Right panel: the system will most likely settle down to an equilibrium
the electron is localised on the thymine. In both cases the excess configuration with the excess electron occupying the SOMO
electron is stabilized by a proton transferred from a neighbouring
glycine (red circles). Reproduced from [147] with permission from and the DNA fragment remaining intact. In other words, the
the Royal Society of Chemistry. system will, upon attaching a very low energy electron to the
DNA strand, relax into the ground state for the anion. It is
ionising the surrounding medium and generating reactive sec- important to note, however, that the work we have presented
ondary species that go on to attack DNA. Which one is more herein has been performed on single DNA bases and on sin-
important is subject to debate. It is estimated that two thirds gle nucleotides. It is reasonable to suppose that these conclu-
of the damage can be attributed to indirect events [11, 12]. sions would also hold for longer DNA fragments but it is also
A thorough discussion of the role of of holes in direct DNA important to remember that dealing with these larger systems
damage events can be found in [21]. Here we focused on indi- introduces new questions. For example: if only one of the
rect damage only. Within this context, the two main points to strands in a double-stranded geometry were to break, the two
elucidate are: how does ionisation of the medium cause DNA ends of the broken strand are prevented from moving apart
damage, and which are the most relevant ionisation products by the hydrogen-bonds to the complementary strand. Whether
in this respect. The two main ionisation products that are gen- or not this is one of the mechanisms that is used in the cell to
erated by the incident radiation are low-energy electrons and repair damaged DNA remains unclear and could be a topic for
radicals. There are many interesting open questions around further investigation.
radicals, in particular their diffusion properties in the medium The previous paragraph may make it sound like the simu-
and the mechanisms via which they interact with DNA and lations presented in this review have shown that, in aqueous
cause damage [148]. Radical interactions with DNA are not environment and at room temperature, strand breaking can-
only relevant in relation to radiation damage, but also central not be triggered by DEA processes involving very low energy
to metabolic damage, thus constituting an interesting avenue electrons. It is important to remember, however, that strand
for research in its own right. breaking reactions can occur in ground state anions and that
In this review we have limited ourselves to the study of the the mechanism for such reactions relies on thermal activation.
interactions of low-energy electrons with DNA, focusing on If this mechanism involving electron capture and relaxation to
the endpoint effects of electron irradiation and leaving aside the the ground state, followed by an activated process in which a
so-called physical stage that describes the ionisation process weakened bond is broken, is dominant, then the rate at which
and the inelastic transport of electrons through the medium. strands are broken should be dependent on temperature.
We have looked in detail at what happens when electrons are In addition to studying the after effects of DEA we have
captured resonantly by DNA nucleobases, nucleotides, and also analysed electron affinities for DNA fragments in the gas-
larger fragments. In particular, we have described the theor phase and in the presence of an aqueous environment that was
etical and experimental investigations that have shown that modelled both explicitly and using a polarisable continuum.
resonances live long enough to form a transient negative ion We concluded from these studies that solvation increases the
that dissociates via the mechanism known as dissociative elec- electron affinity and that hydrogen-bonding plays a crucial
tron attachment, and that this mechanism is particularly preva- role. We have also modelled bond breaking reactions that
lent when DNA is in an aqueous environment. To examine the occur on the energy landscape of the ground state anion in a
dynamics after a DEA event we have used a model in which variety of different situations. These studies used constrained
the excess electron energy is instantaneously transferred to a first-principles molecular dynamics simulations to compute
specific bond as vibrational energy. The main conclusion of free energy profiles and activation barriers. These calculated
these studies is that bond dissociation energies calculated in barriers were on the order of 20 kcal mol−1, but they can
gas phase cannot be used to estimate the likelihood that a bond increase or decrease depending on the environment. In par
will break when the base or nucleotide is solvated. When mol- ticular, protonation of the base tends to raise the barriers and
ecules are solvated, breaking bonds requires a significantly thus protects nucleotides against strand breaks. Future work
greater input of energy as the surrounding water molecules in this direction would involve simulating larger fragments
can reflect energy back into the solute through caging effects. such as polynucleotides and double-stranded small fragments
26
in order to determine if there are particular sequences of [10] Lodge M, Pijls-Johannesma M, Stirk L, Munro A J,
nucleobases that favour or inhibit strand breaking reactions. Ruysscher D D and Jefferson T 2007 Radiother. Oncol.
83 110–22
There are then open questions regarding how double strand
[11] Alizadeh E and Sanche L 2012 Chem. Rev. 112 5578–602
breaks occur and what role the amino acids of the histones [12] Jonah C and Rao B 2001 Radiation Chemistry: Present Status
play in this process. We have shown that amino acids protect and Future Trends (Studies in Physical and Theoretical
nucleobases by protonating them or by scavenging electrons. Chemistry) (Amsterdam: Elsevier)
Whether or not this remains the case for nucleotides and larger [13] Nikjoo H, Uehara S, Emfietzoglou D and Brahme A 2008 New
J. Phys. 10 075006
fragments, and in models with a more accurate representation
[14] Kallikragas D T, Plugatyr A Y and Svishchev I M 2014
of chromatin is an open question, however. J. Chem. Eng. Data 59 1964–9
In short, our understanding of the damage caused to DNA [15] Surdutovich E, Yakubovich A V and Solov’yov A V 2013
by low-energy electrons has come a long way since the first Sci. Rep. 3 1289
experiments in Sanche’s group and new experiments and [16] de Vera P, Mason N J, Currell F J and Solov’yov A V 2016
Eur. Phys. J. D 70 183
simulations continue to shed light into the matter. We have,
[17] Lewis F D, Liu X, Liu J, Miller S E, Hayes R T and
however, only really dipped our toe into this vast problem. A Wasielewski M R 2000 Nature 406 51–3
colossal amount of work is still required until we get a clear [18] Giese B, Amaudrut J, Kohler A K, Sporran M and Wessely S
understanding into how DNA damage by LEE is achieved in 2001 Nature 412 318–20
physiological environments, in part because the cell contains [19] Conwell E M and Rakhmanova S V 2000 Proc. Natl Acad.
Sci. USA 97 4556–60
enzymatic machinery that can interact with broken DNA and
[20] Shao F, O’Neill M A and Barton J K 2004 Proc. Natl Acad.
repair it [149]. In the meantime, this review suggests a number Sci. USA 101 17914–9
of open research avenues. Hopefully these will be explored in [21] Yun B H, Lee Y A, Kim S K, Kuzmin V, Kolbanovskiy A,
the near future. Dedon P C and Shafirovich V 2007 J. Am. Chem. Soc.
129 9321–32
[22] Livshits G I et al 2014 Nat. Nanotechnol. 9 1040–6
Acknowledgments [23] Pimblott S M and La Verne J A 2007 Radiat. Phys. Chem.
76 1244–7
[24] Steenken S 1989 Chem. Rev. 89 503–20
We thank Maeve Smyth, Ilya Fabrikant, Lila Bouëssel du [25] Kumar A, Adhikary A, Shamoun L and Sevilla M D 2016
Bourg, Amy Williamson and Declan Scullion for their contrib J. Phys. Chem. B 120 2115–23
utions to parts of this work. We also thank Michael Ferguson for [26] von Sonntag C 1987 The Chemical Basis of Radiation Biology
help with the chemical drawings, and Pablo de Vera for creat- (London: Taylor and Francis)
[27] Breen A P and Murphy J A 1995 Free Radical Biol. Med.
ing figure 4 of the nucleosome. We have had numerous useful
18 1033–77
discussions with Fred Currell, Anil Kumar, Amitava Adhikary, [28] Alizadeh E, Sanz A G, Garcia G and Sanche L 2013
Ilya Fabrikant and Leon Sanche for which we are grateful. J. Phys. Chem. Lett. 4 820–5
Lastly we received computational support from the UK national [29] Haxton D J, Zhang Z, Meyer H D, Rescigno T N and
high performance computing service, ARCHER via the UKCP McCurdy C W 2004 Phys. Rev. A 69 062714
[30] Sanche L 2009 Radical and Radical Ion Reactivity in Nucleic
consortium and funded by EPSRC grant ref EP/K013564/1.
Acid Chemistry (Hoboken, NJ: Wiley) pp 239–94
[31] Arumainayagam C R, Lee H, Nelson R B, Haines D R and
Gunawardane R 2010 Surf. Sci. Rep. 65 1–44
ORCID iDs
[32] Baccarelli I, Bald I, Gianturco F, Illenberger E and Kopyra J
2011 Phys. Rep. 508 1–44
Jorge Kohanoff https://orcid.org/0000-0002-8237-7543 [33] Alizadeh E, Orlando T and Sanche L 2015 Annu. Rev. Phys.
Gareth A Tribello https://orcid.org/0000-0002-4763-9317 Chem. 66 379–98
[34] Kumar A and Sevilla M 2012 Handbook of Computational
Chemistry vol III: Applications—Biomolecules ed M Shukla
References and J Leszczynski (Netherlands: Springer) pp 1215–56
[35] Caron L and Sanche L 2012 Low-Energy Electron Scattering
from Molecules, Biomolecules and Surfaces ed P Čársky
[1] Boudaiffa B, Cloutier P, Hunting D, Huels M A and Sanche L and R Čurik (Boca Raton, FL: Taylor and Francis)
2000 Science 287 1658–60 [36] Gu J, Leszczynski J and Schaefer I H 2012 Chem. Rev.
[2] Koltzoff N K 1928 Biol. Zentralblatt 48 345–69 112 5603–40
[3] Avery O T, Macleod C M and McCarty M 1944 J. Exp. Med. [37] Cobut V, Frongillo Y, Patau J P, Goulet T, Fraser M J and
79 137–8 Jay-Gerin J P 1998 Radiat. Phys. Chem. 51 229–43
[4] Watson J D and Crick F H 1953 Nature 171 737–8 [38] Singh N P, McCoy M T, Tice R R and Schneider E L 1988
[5] Friedberg E C 2003 Nature 421 436–40 Exp. Cell Res. 175 184–91
[6] Belau L, Wilson K R, Leone S R and Ahmed M 2007 J. Phys. [39] Luo X, Zheng Y and Sanche L 2014 J. Chem. Phys.
Chem. A 111 7562–8 140 155101
[7] Dawley M M, Tanzer K, Cantrell W A, Plattner P, [40] Russo N, Toscano M and Grand A 2000 J. Comput. Chem.
Brinkmann N R, Scheier P, Denifl S and Ptasinska S 2014 21 1243–50
Phys. Chem. Chem. Phys. 16 25039–53 [41] Barrios R, Skurski P and Simons J 2002 J. Phys. Chem. B
[8] Page R H, Larkin R, Yhen Y R and Lee Y T 1988 J. Chem. 106 7991–4
Phys. 88 2249 [42] Martin F, Burrow P D, Cai Z L, Cloutier P, Hunting D and
[9] IAEA: Radiation in Everyday Life. www.iaea.org/ Sanche L 2004 Phys. Rev. Lett. 93 068101
Publications/Factsheets/English/radlife. Accessed: [43] Huels M, Hahndorf I, Illenberger E and Sanche L 1998
13–03-2017 J. Chem. Phys. 108 1309–12
27
[44] Fisher D J and Zhang C 1994 J. Appl. Phys. 76 606–8 [80] van Faassen M, Wasserman A, Engel E, Zhang F and
[45] Christophorou L and Olthoff J 2012 Fundamental Electron Burke K 2007 Phys. Rev. Lett. 99 043005
Interactions with Plasma Processing Gases (Physics of [81] Rizzi V, Todorov T N, Kohanoff J J and Correa A A 2016
Atoms and Molecules) (New York: Springer) Phys. Rev. B 93 024306
[46] Denifl S, Ptasinska S, Cingel M, Matejcik S, Scheier P and [82] Trani F, Scalmani G, Zheng G, Carnimeo I, Frisch M J and
Märk T 2003 Chem. Phys. Lett. 377 74–80 Barone V 2011 J. Chem. Theory Comput. 7 3304–13
[47] Hanel G, Gstir B, Denifl S, Scheier P, Probst M, Farizon B, [83] Negre C F A, Fuertes V C, Oviedo M B, Oliva F Y and
Farizon M, Illenberger E and Märk T D 2003 Phys. Rev. Sánchez C G 2012 J. Phys. Chem. C 116 14748–53
Lett. 90 188104 [84] Correa A A, Kohanoff J, Artacho E, Sánchez-Portal D and
[48] Gohlke S, Abdoul-Carime H and Illenberger E 2003 Caro A 2012 Phys. Rev. Lett. 108 213201
Chem. Phys. Lett. 380 595–9 [85] Horsfield A P, Bowler D R, Fisher A J, Todorov T N and
[49] Abdoul-Carime H, Gohlke S and Illenberger E 2004 Sánchez C G 2005 J. Phys.: Condens. Matter 17 4793
Phys. Rev. Lett. 92 168103 [86] McEniry E J et al 2010 Eur. Phys. J. B 77 305–29
[50] Ptasinska S, Denifl S, Grill V, Mark T D, Illenberger E and [87] Rajam A K, Raczkowska I and Maitra N T 2010 Phys. Rev.
Scheier P 2005 Phys. Rev. Lett. 95 093201 Lett. 105 113002
[51] Ptasinska S, Denifl S, Scheier P, Illenberger E and Mark T [88] Tully J C 1990 J. Chem. Phys. 93 1061–71
2005 Angew. Chem., Int. Ed. 44 6941–3 [89] Rizzi V, Todorov T N and Kohanoff J J 2017 Sci. Rep.
[52] Ibanescu B C and Allan M 2009 Phys. Chem. Chem. Phys. 7 45410
11 7640–8 [90] Hutter J, Iannuzzi M, Schiffmann F and VandeVondele J
[53] Čársky P and Čurik R (ed) 2016 Low-Energy Electron 2014 Wiley Interdiscip. Rev.: Comput. Mol. Sci. 4 15–25
Scattering from Molecules, Biomolecules and Surfaces [91] Smyth M and Kohanoff J 2011 Phys. Rev. Lett. 106 238108
(Boca Raton, FL: CRC Press) [92] Kočišek J, Pysanenko A, Farnik M and Fedor J 2016 J. Phys.
[54] Tennyson J 2010 Phys. Rep. 491 29–76 Chem. Lett. 7 3401–5
[55] Fabrikant I 1991 Phys. Rev. A 43 3478–86 [93] Berg J M, Tymoczko J L and Stryer L 2002 Biochemistry 5th
[56] Domcke W 1991 Phys. Rep. 208 97–188 edn (New York: W.H. Freeman) (International Version)
[57] McCurdy C, Isaacs W, Meyer H D and Rescigno T 2003 [94] Kumar A and Sevilla M D 2007 J. Phys. Chem. B
Phys. Rev. A 67 042708–19 111 5464–74
[58] Caron L and Sanche L 2003 Phys. Rev. Lett. 91 [95] Kumar A and Sevilla M 2009 ChemPhysChem 10 1426–30
[59] Caron L, Sanche L, Tonzani S and Greene C H 2008 [96] Steenken S 1992 Free Radical Res. Commun. 16 349–79
Phys. Rev. A 78 042710 [97] Liu B, Nielsen S B, Hvelplund P, Zettergren H,
[60] Gallup G and Fabrikant I 2011 Phys. Rev. A 83 012706 Cederquist H, Manil B and Huber B A 2006
[61] Nesbet R 1981 Phys. Rev. A 24 1184–93 Phys. Rev. Lett. 97 133401
[62] Hotop H, Ruf M W, Allan M and Fabrikant I 2003 Adv. At. [98] Gu J, Xie Y and Schaefer H F 2007 Nucl. Acids Res.
Mol. Opt. Phys. 49 85–216 35 5165–72
[63] Smyth M, Kohanoff J and Fabrikant I 2014 J. Chem. Phys. [99] Gu J, Xie Y and Schaefer H III 2010 J. Phys. Chem. B
140 184313 114 1221–4
[64] Bald I, Kopyra J and Illenberger E 2006 Angew. Chem., Int. [100] Haranczyk M, Gutowski M, Li X and Bowen K 2007 J. Phys.
Ed. 45 4851–5 Chem. B 111 14073–6
[65] Baccarelli I, Gianturco F, Grandi A, Sanna N, Lucchese R, [101] Kim S, Wheeler S E and Schaefer H F 2006 J. Chem. Phys.
Bald I, Kopyra J and Illenberger E 2007 J. Am. Chem. Soc. 124 204310
129 6269–77 [102] Kim S and Schaefer H 2006 J. Chem. Phys. 125 144305
[66] Bald I, Flosadettir H D, Kopyra J, Illenberger E and [103] Dedikova P, Neogrady P and Urban M 2011 J. Phys. Chem.
Ingelfsson O 2009 I. J. Mass Spectrom. 280 190–7 A 115 2350–8
[67] Abdoul-Carime H, Gohlke S, Fischbach E, Scheike J and [104] Smets J, McCarthy W and Adamowicz L 1996 Chem. Phys.
Illenberger E 2004 Chem. Phys. Lett. 387 267–70 Lett. 256 360–9
[68] Ptasinska S, Denifl S, Gohlke S, Scheier P, Illenberger E and [105] Gu J, Wang J and Leszczynski J 2006 J. Am. Chem. Soc.
Mark T 2006 Angew. Chem., Int. Ed. 45 1893–6 128 9322–3
[69] Konig C, Kopyra J, Bald I and Illenberger E 2006 Phys. Rev. [106] Gu J, Wang J and Leszczynski J 2010 Nucl. Acids Res.
Lett. 97 018105 38 5280–90
[70] Burrow P, Gallup G and Modelli A 2008 J. Phys. Chem. A [107] Gu J, Wang J and Leszczynski J 2011 J. Phys. Chem. B
112 4106–13 115 14831–7
[71] Bald I, Dabkowska I and Illenberger E 2008 Angew. Chem., [108] Schyman P, Laaksonen A and Hugosson H 2008 Chem. Phys.
Int. Ed. 47 8518–20 Lett. 462 289–94
[72] Bao X, Wang J, Gu J and Leszczynski J 2006 Proc. Natl Acad. [109] Li Z, Cloutier P, Sanche L and Wagner J R 2010 J. Am.
Sci. USA 103 5658–63 Chem. Soc. 132 5422–7
[73] Simons J 2006 Acc. Chem. Res. 39 772–9 [110] Dinh P M et al 2017 Nanoscale Insights into Ion-Beam
[74] Dora A, Bryjko L, van Mourik T and Tennyson J 2012 J. Phys. Cancer Therapy ed A V Solov’yov (Cham: Springer)
B: At. Mol. Opt. Phys. 45 175203 pp 277–309
[75] Dora A, Bryjko L, Van Mourik T and Tennyson J 2012 [111] Zheng Y, Wagner J and Sanche L 2006 Phys. Rev. Lett.
J. Chem. Phys. 136 024324 96 208101
[76] Sieradzka A and Gorfinkiel J 2017 J. Chem. Phys. under [112] Ptasinska S and Sanche L 2007 Phys. Chem. Chem. Phys.
review 9 1730–5
[77] Fabrikant I, Eden S, Mason N and Fedor J 2017 Adv. At. Mol. [113] Zheng Y, Cloutier P, Hunting D, Sanche L and Wagner J
Opt. Phys. 66 545–657 2005 J. Am. Chem. Soc. 127 16592–8
[78] Görling A 1996 Phys. Rev. A 54 3912–5 [114] Gu J, Xie Y and Schaefer H F III 2006 J. Am. Chem. Soc.
[79] Kohanoff J 2006 Electronic Structure Calculations for Solids 128 1250–2
and Molecules: Theory and Computational Methods [115] Berdys J, Anusiewicz I, Skurski P and Simons J 2004
(Cambridge: Cambridge University Press) J. Phys. Chem. A 108 2999–3005
28
[116] Svozil D, Jungwirth P and Havlas Z 2004 Collect. Czech. [135] Gu J, Wang J and Leszczynski J 2010 ChemPhysChem
Chem. Commun. 69 1395–428 11 175–81
[117] Schiedt J, Weinkauf R, Neumark D and Schlag E 1998 Chem. [136] Wang C R, Nguyen J and Lu Q 2009 J. Am. Chem. Soc.
Phys. 239 511–24 131 11320–2
[118] Desfrancois C, Periquet V, Bouteiller Y and Schermann J [137] Uhlig F, Marsalek O and Jungwirth P 2012 J. Phys. Chem.
1998 J. Phys. Chem. A 102 1274–8 Lett. 3 3071–5
[119] Colson A O, Besler B and Sevilla M 1992 J. Phys. Chem. [138] Smyth M and Kohanoff J 2012 J. Am. Chem. Soc.
96 9787–94 134 9122–5
[120] Roca-Sanjuan D, Merchan M, Rubio M and [139] McAllister M, Smyth M, Gu B, Tribello G A and Kohanoff J
Serrano-Andres L 2008 J. Chem. Phys. 129 095104 2015 J. Phys. Chem. Lett. 6 3091–7
[121] Riley K, Op’t Holt B and Merz K Jr 2007 J. Chem. Theory [140] McAllister M 2016 Computational modelling of radiation
Comput. 3 407–33 damage to DNA PhD Thesis Queen’s University Belfast
[122] Wesolowski S S, Leininger M L, Pentchev P N and [141] Steenken S, Telo J, Novais H and Candeias L 1992 J. Am.
Schaefer H F III 2001 J. Am. Chem. Soc. 123 4023–8 Chem. Soc. 114 4701–9
[123] Sarmah P and Deka R C 2008 Mol. Simul. 34 879–85 [142] Cauet E, Bogatko S, Livin J, De Proft F and Geerlings P
[124] Becke A 1993 J. Chem. Phys. 98 5648–52 2013 J. Phys. Chem. B 117 9669–76
[125] Lee C, Yang W and Parr R 1988 Phys. Rev. B 37 785–9 [143] Ito T, Baker S, Stickley C, Peak J and Peak M 1993
[126] Zhao Y, Schultz N and Truhlar D 2006 J. Chem. Theory Int. J. Radiat. Biol. 63 289–96
Comput. 2 364–82 [144] Szyperska A, Gajewicz A, Mazurkiewicz K, Leszczynski J
[127] Zhao Y and Truhlar D 2008 Theor. Chem. Acc. 120 215–41 and Rak J 2011 Phys. Chem. Chem. Phys.
[128] Smyth M 2012 A computational study of radiation damage to 13 19499–507
DNA PhD Thesis Queen’s University Belfast [145] Solomun T and Skalicky T 2008 Chem. Phys. Lett.
[129] Chen E and Chen E 2000 J. Phys. Chem. B 104 7835–44 453 101–4
[130] Matthew J and Richards F 1984 Biopolymers 23 2743–59 [146] Ptasinska S, Li Z, Mason N J and Sanche L 2010
[131] Sponer J, Burda J, Sabat M, Leszczynski J and Hobza P 1998 Phys. Chem. Chem. Phys. 12 9367–72
J. Phys. Chem. A 102 5951–7 [147] Gu B, Smyth M and Kohanoff J 2014 Phys. Chem. Chem.
[132] Mennucci B 2012 Wiley Interdiscip. Rev.: Comput. Mol. Sci. Phys. 16 24350–8
2 386–404 [148] Wu Y, Mundy C J, Colvin M E and Car R 2004 J. Phys.
[133] Ptasinska S and Sanche L 2007 Phys. Rev. E 75 031915 Chem. A 108 2922–9
[134] Schyman P and Laaksonen A 2008 J. Am. Chem. Soc. [149] Banerjee A, Yang W, Karplus M and Verdine G L 2005
130 12254–5 Nature 434 612–8
29

Topical JPCM Published

Uploaded by

Copyright:

Available Formats

Topical JPCM Published

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Topical JPCM Published

Uploaded by

Copyright:

Available Formats

Home Search Collections Journals About Contact us My IOPscience

2017 J. Phys.: Condens. Matter 29 383001

Please note that terms and conditions apply.

You may also be interested in:

Ultrafast molecules and clusters

Applications of large-scale density functional theory in biology

Theoretical description of protein field effects on electronic excitations of biological

J. Phys.: Condens. Matter 29 (2017) 383001 (29pp) https://doi.org/10.1088/1361-648X/aa79e3

Interactions between low energy electrons

Received 18 January 2016, revised 28 May 2017

1361-648X/17/383001+29$33.00 1 © 2017 IOP Publishing Ltd Printed in the UK

DNA was discovered in 1869 by Friedrich Miescher. It then

1.1. DNA structure

viable any longer because the genetic material is damaged,

1.2. DNA damage 1.3. Radiation damage of DNA

The second of these two reactions involves the radical water

room temperature. Understanding DNA damage under these

1.4. DNA damage by low-energy electrons

LEE generated by water radiolysis can be captured by DNA

Figure 9. Thymine base with ring atoms numbered. When a 1 eV

of DNA. At higher energies the TNI formed are Feshbach

sphere with external incoming and outgoing waves, which are

3. Dynamics ensuing DEA: from the gas to the

In the absence of an adequate methodology for describing

with the Thymine at all possible positions involving N and

environ­ment we simulated the post-DEA dynamics in both

breaking of this bond. In the non-hydrogen bonded case, the

for the capture cross section. Gas-phase calculations indicate

Theory level AEA (eV) Functional AEA (eV)

post-HF, coupled cluster (CCSD) methods, that the AEA of

4.3.2. Solvation effects on dissociation. The electron capture

study suggested that protonation stabilised the anion and that,

There is a considerable amount of evidence suggesting that

These processes bounce energy between the solute and the

You might also like

environment we simulated the post-DEA dynamics in both