Biophysical
Elsevier
Chemistry,
33 (1989) 91-98
91
BPC 01347
Linkage effects in a model for cell survival after radiation
Enrico Di Cera, Francesco Andreasi Bassi and Giuseppe Arcovito
Istituio di Fisico, Universitd Cattolica, Largo F. Vito I, 00168 Roma, Italy
Received 25 October 1988
Accepted 9 December 1988
‘l%ermodynamics;
Radiation
damage;
Cell death
A thermodynamic
treatment
for the effects of radiation
on cell survival is proposed.
The treatment
is an extension of the
linear-quadratic
model (K.H. Chadwick and H.P. Lee&outs,
Phys. Med. Biol. 13 (1973) 78) following the principles of linkage
thermodynamics
(E. Di Cera, S.J. Gill and J. Wyman, Proc. Natl. Acad. Sci. U.S.A. 85 (1988) 5077). Linkage effects between
chemical binding to DNA and radiation action are considered, along with the synergism between different types of radiations. A
simple mathematical
condition is found for the additivity of radiation doses that result in an isoeffect. The resolvability of the model
parameter is investigated by simulations and statistical analysis of the distributions
obtained.
1. Introduction
The quantitative analysis of the effects of radiations on living cells has resulted in several attempts todevelop suitable’models for the description of the dose-response curves observed experimentally [1,2]. It is generally accepted that among
the possible molecular lesions induced by radiation action on living cells, those involving the
DNA molecule are of critical importance. These
lesions arise from radicals produced by the passage of the radiation through the aqueous medium
surrounding the DNA. In particular, the doublestrand break (dsb) of the DNA molecule represents a most critical lesion which easily leads to
cell death [3,4]. On the other hand, the singlestrand break (ssb) is a molecular lesion which is
efficiently repaired [5] and is not of great importance for the biological effect. Consistent with
these experimental facts, Chadwick and Leer&outs
[6,7] have proposed a model, referred to as the
linear-quadratic (LQ) model, which relates cell
survival, S, to the number of dsbs induced by
radiation action in the DNA molecule. This number is assumed to be a linear-quadratic function of
the dose of radiation, D, so that
S=exp(-aD-pD*)
where (Y and j3 are two parameters to be determined experimentally.
The LQ model postulates that the dsb of the
DNA molecule can be produced according to two
different mechanisms. The first is breakage induced by a single ionizing particle and is proportional to D. The second is breakage induced by
two separate ionizing particles and is proportional
to D2. Combination of these two possible meehanisms of cell damage yields eq. 1 for the probability of cell survival, after assuming that the lethal
events follow the Poisson distribution [7].
The LQ model has successfully been applied to
the analysis of a substantial body of experimental
data [7]. Its beauty stems from both the mathematical simplicity and the underlying physical
mechanisms invoked to explain the observed biological effects. The model parameters (Y and fi
have been measured for several cell types and
Correspondence
address: E. Di Cera, Istituto di Fisica,
versita Cattolica, Large F. Vito 1, 00168 Roma, Italy.
Uni-
0301-4622/89/503.50
B.V. (Biomedical
8 1989 Elsevier Science Publishers
(1)
Division)
92
E. Di Cera et nl./Linkage
effectsm CImodelfor cell swvival after radiation
tissues and, in some cases, under a variety of
experimental conditions. This provides a considerable data base which allows for a deeper understanding of both the molecular theory and the
significance of its physical parameters. It is the
purpose of this paper to give the LQ model a
broader theoretical basis, using considerations derived from the law of mass action. The approach
we take here is cast within the framework of
linkage thermodynamics [8,9] and is focussed on
the phenomenological treatment of linkage effects.
These effects have been documented experimentally by studying, for example, the influence of
ionic strength on cell survival after radiation [lo],
or the synergistic action of radiations of different
nature [ll]. However, their possible relevance in
connection with a more general formulation of the
LQ model has not been recognized. It is therefore
of relevance to develop such a general formulation
for the description of reciprocal effects in the
interaction of radiation with living cells. The
strength of this approach is revealed by the derivation of linkage relations that have an immediate
practical application to the analysis of experimental data. These relations broaden our understanding of the detailed molecular mechanisms of radiation action and provide a quantitative basis for
further theoretical investigation of the subject.
Along with the theoretical extension of the LQ
model, we consider here the problem of the resolvability of the model parameters and their correlation.
2. Theory
We start from the basic postulate of the LQ
model and consider the elementary effect responsible for cell death, the dsb of the DNA molecule.
This effect can be produced in two different ways:
the first one is due to a single ionizing particle and
is proportional to the dose D. The second one is
produced by two independent hits and is proportional to D*. In these basic assumptions we recognize a parallel with the law of mass action, in the
sense that the relation between the dose D and the
effect observed, i.e., the number of ionizing particles producing a dsb, can be formulated as in the
case of chemical binding phenomena in a biological macromolecule. In this respect the dose D can
be considered to play the same role as the activity
of a ligand. With such a parallel in mind we write
down a partition function, 2, as follows
Z=l+aD+bD*
(2)
Each term in the partition function is proportional
to a given configuration relative to the elementary
effect being considered in the interaction of the
radiation with the DNA. The first term refers to
the absence of dsbs, and hence of biological
damage. This term may include ssbs followed by
repair. The second refers to a dsb due to a single
ionizing particle, and the third one refers to a dsb
produced by two separate ionizing particles.
From the partition function we compute by
differentiation the average number of ionizing
particles, n, that have produced a dsb at a given
dose, D, as follows
dlnZ
-=
aD -I 2bD2
d In D
l+aD+bDZsn
(3)
The coefficients a and b can be given a mass law
interpretation as ‘affinity constants’ for the occurrence of a dsb induced in the two alternative ways.
The coefficient a is the inverse of the dose which
gives 50% of the ionizing particles that can induce
a dsb by a single hit. The coefficient b is the
inverse of the dose squared which gives 50% of the
ionizing particles that can induce a dsb by two
separate hits. Eq. 3 thus gives the average number
of ‘lethal ionizing particles’ per dsb. if we assume
that the lethal events are all independent, then the
total number of lethal ionizing particles per DNA
molecule, ZV,is given by
N=m
aD + 2bD2
1+aD+bD2
where m is the number of base-pairs per DNA
molecule. One sees from eq. 4 that when b = 0 the
maximum number of lethal ionizing particles per
DNA equals the number of base-pairs, since they
all produce a dsb by a single hit. On the other
hand, for positive values of the coefficient b the
E. Di Cera et al./ Linkage effects in a model for cell survival after radiation
maximum value of N is 2m, since dsbs induced by
separate hits become predominant at high doses.
Since n is very small at doses that are of
interest in the study of eukaryotic cells [12], the
value of N can be approximated by
N=aD-@D’
(5)
The coefficients (Yand fi are now combinations of
the constants a and b and the number of basepairs m per DNA molecule. Experimental determination of a and p shows that they are of the
order of 10-l Gy-’ and lo-’ Gy-‘, respectively.
Since m is of the order of 3 X lo9 in eukaryotic
cells, we conclude that the coefficients a and b
must be of the order of lo-” Gy-’ and lo-”
Gye2, respectively. Therefore, in the dose range
from 0 to 30 Gy, which is the one usually employed in experimental studies of radiation action,
the value of 2 is practically equal to 1 and n is
indeed negligible. This justifies the use of eq. 5 for
practical purposes. Saturation effects, that are unlikely to occur at the doses used experimentally
[12], can however be treated with the use of eq. 4,
which is the exact mathematical form relating the
dose D to the number of lethal ionizing particles
N.
Although the occurrence of a dsb is intrinsically a rare event (see eq. 4), nevertheless the
number of possible dsbs per DNA molecule is
almost infinite (about 3. lo’), which makes it
convenient to assume that the number of lethal
ionizing particles per DNA, at a given dose, follows the Poisson distribution. This is a straightforward application of basic principles of statistics
[13]. The probability that there are 0 lethal ionizing particles per DNA gives the probability of cell
survival, S, at a given dose, D, as
S=exp(-N)=exp(-aD-/3D*)
(6)
which is identical to the expression for cell survival
of the LQ model [6,7].
This result, obtained from considerations based
on the law of mass action, provides the theoretical
basis for an extension of the LQ model along the
principles of linkage thermodynamics [8,9], where
the quantitative description of linkage effects
dominates the picture.
93
3. Linkage effects
The probability of cell survival as a function of
the dose of radiation can be affected by several
factors. Some of them are linked to the cell cycle,
mitosis being the most sensitive phase to radiation
action [14,15]. Other factors that reduce cell
survival are exposure to high oxygen partial pressures [la], or lowering the ionic strength of the
medium [lo]. From the point of view of linkage
thermodynamics, which is the one we are concerned with here, any substance which binds to
the DNA molecule may affect the occurrence of a
dsb and vice versa. The mutual interference between radiation damage and the binding of substances to DNA can be considered as a linkage
effect that can be explored experimentally,
Starting from the partition function (eq. 2) we
describe the effect of ligand binding to DNA and
radiation action as follows
Z=
fi Aojxj+
j=O
5 A,,x’D+
j=O
i
A2jxjD2
(7)
j=O
Here the A parameters are mass law coefficients
describing both ligand binding of an arbitrary
substance x, whose activity is x, and the effect of
radiation. The constant A, is the overall equilibrium constant for the reaction BP +jX = BPX,
(BP = base-pair) when no dsb has occurred. The
constants Alj and Azj describe the same reaction
when a dsb has occurred by a single hit or two
separate ones, respectively. The parameter p is the
number of binding sites for ligand X per base-pair.
The number of ligand molecules bound per DNA,
X, at a given dose D, is
X=m
(alnz 1
alnx
a
.
and the number of lethal ionizing particles per
DNA, at a given ligand activity x, is given by
allIz
NEm
(alnD
)
x
From the relations above it is easy to derive
linkage relations such as
E. Ri Cera et ai./Limkage effects in a model forcell survival after radiation
94
(11)
that give a measure of the mutual interaction
between chemical binding and radiation effects.
The most important
implication
of the existence of a linkage between chemical binding and
radiation effects is the possibility of expressing the
coefficients (Yand j3 of the LQ model as a function of the ligand activity x as follows
I
P
P
(Y =
m
C Al,xJ/
02)
j=O
j-0
j-0
C AojXj
.
j=o
.
(13)
Hence
d In (Y
-=
dln x
Xi - X, = SX,,
d In fl
d
=x,-x,=sx,,
04)
(15)
The relations above have an immediate practical
application.
The dependence of In (Y and In p on
the logarithm of the ligand activity x gives the
change in ligand molecules bound to DNA upon
induction of a dsb. The term 8X,, is a measure of
the linkage when the dsb is induced by a single
ionizing particle, while SX,, is the corresponding
linkage for a dsb induced by two separate ionizing
particles.
An example of the application of eqs. 14 and
15 is reported in fig. 1. Cell survival curves obtained at different Na+ concentrations
[lo] show a
significant change of the two coefficients cr and j3.
The slope of the line interpolating
the experimental points yields - 0.2 mbl Naf per dsb in the case
of a single ionizing particle, and -0.44 mol Na+
per dsb in the case of two ionizing particles.
Interestingly,
the two linkage coefficients
SX,,
and 8X,, seem to scale with the number of lethal
ionizing particles involved in a dsb, which suggests
that 0.2 mol Na+ are released per lethal ionizing
particle. One sees how the linkage approach developed here can be used to unravel the quantitative
-I.5
-1
-0.5
LOQ
0
as
[No +]
Fig 1. Linkage between Naf binding and radiation action to
DNA in the case of Chinese hamster cells [lo], illustrated by
the change of the logarithm of (Y(B) and fi (0) as a function of
the logarithm
of Na* activity. One sees that cell survival
increases with increasing Na+ activity. This means that Na+
binding to DNA stabilizes the configuration
with no dsb with
respect to the others (see eq. 7). The two lines are the best fits
to the experimentaL data. The slope of each line P;ves the
number of Na* released when a dsb is induced by a single
ionizing particle (m) or two separate ones (O), as implied by
eqs. 14 and 15. The values are: 8X,, = -0.21 kO.06, and
SX,, = - 0.5Oi 0.03. Note how the ratio 8 X,,/GXIo
is practically equal, within errors, to 2. This suggests that the number
of Na+ released per lethal ionizing particle is constant.
aspects of the mechanism of radiation action. This
approach
is thus of critical importance
in the
analysis of experimental
data revealing a mutual
interference
of chemical binding
and radiation
damage.
4. Synergism as identical linkage
In the description given above concerning linkage effects between chemical binding and radiation action we have shown how experimental
determination
of the two coefficients a and j3 can be
used to evidentiate
chemical
phenomena
that
accompany the occurrence of a dsb. Another class
of linkage effects to be considered here is the one
involving two different types of radiations
that
both lead to dsbs. This phenomenon,
usually
referred to as synergism, is of extreme relevance in
both biological and medical fields [7]. It is here
that the strength and elegance of the approach
E. Di Cera et aL/Linkage
effects m a model for ceil survival after radiation
based on linkage thermodynamics
can be appreciated best. From the standpoint of linkage thermodynamics the synergistic interaction of two radiations of different nature is an example of identical
linkage [9,17,18]. In this case the partition function is given by
Z=1+aD+bD2+sE+tE2+cDE
(16)
where E denotes the dose of the second radiation,
and the coefficients s and t have the same significance as u and b in eq. 2, and refer to the second
type of radiation. The coefficient c refers to the
configuration
where a dsb is produced by two
separate ionizing particles, one for each type of
radiation.
The ‘identical’ nature of this kind of linkage
arises from the fact that where a dsb has occurred
due to the action of one type of radiation, there is
no possibility of producing a dsb by means of the
other type of radiation. In other words, the effects
of the two types of radiations are mutually exclusive, as seen in the binding reactions of oxygen
and carbon monoxide to the heme site of human
hemoglobin [18], which has led to the definition of
identical linkage [17]. In this respect, the identical
linkage is quite different from the linkage described by eq. 7, where the two effects under
consideration,
i.e., ligand binding and radiation
action, are not mutually exclusive. Notwithstanding, they are identically linked, and hence mutually exclusive, the two types of radiation lead to
the same molecular lesion, the dsb. In such a
feature one recognizes their synergistic interaction.
The number of lethal ionizing particles per DNA
due to the first type of radiation can be calculated
as in the case of eqs. 3-5 as follows
and similarly
has
The probability
for the second type of radiation
one
of cell survival is then given by
95
which is identical to the expression derived by
Chadwick and Leenhouts [7] without taking into
account considerations
based on the law of mass
action. Eq. 19 shows that S is a function of the
sum N + Q, and that the two types of radiation
arc synergistic in producing cell death.
From the partition
function
one can derive
linkage relations such as eqs. 10 and 11. There is,
however, a most important
aspect in connection
with the synergistic interaction of the two types of
radiation. The ratio N/Q gives a measure of the
partitioning
between lethal ionizing particles due
to the two types of radiation with respect to the
DNA molecule. A partition law can be formulated
as follows
N/Q = @/E
(20)
i.e., the ratio of lethal ionizing particles due to the
two types of radiation is proportional
to the ratio
of the respective doses times a constant factor @,
the partition
coefficient.
This law is the exact
parallel of the Haldane law [19] for the partitioning of two identically linked ligands such as oxygen
and carbon monoxide to human hemoglobin [18].
The partition
coefficient is equal to the ratio of
the doses E/D when N = Q. The validity of the
partition law given above implies that the survival
curves obtained as a function of the logarithm of
the dose of the two types of radiations must be
parallel, the displacement
of the two curves being
constant and equal to the logarithm of 9. This
simple condition
can readily be tested from the
analysis of experimental data. Failure of the partition law necessarily implies that + is not constant.
An important consequence of the partition law
is that if 4 is indeed constant then the sum N + Q
is a function of E + $10 only. This means that in
the domain where the partition law, eq. 20, holds
the doses of the two types of radiation for which
N + Q is constant are additive. Demonstration
is
as follows. From the partition function, eq. 16, it
results that [9]
(wN+e=- (iy,,,_,,,
(21)’
S=exp(-N-Q)
=exp(-aD-j3D2-vE-rE2-2pDE)
(19)
This is a most important
identically linked ligands
relation in the theory of
[9,17], and is even more
96
E. Di Cera et aI./ Linkage effects in a model for ceil survival after radiation
relevant in connection with the phenomenon of
synergism because it describes an isoeffect. In
other words, since S is a function of N + Q only,
when this sum is constant so is the effect. Therefore; the partial derivative on the left-hand side of
eq. 21 gives the dependence of In E on ln D
which results in an isoeffect. This partial derivative is equal to the change of Q with respect to N
which keeps the ratio D/E constant. Now, if the
partition law, eq. 20, holds, then the right-hand
side of eq. 21 is simply @D/E, so that
(22)
Hence,
i.e., the sum E + ~$0 is a function of N + Q only
and, conversely, the sum N + Q is a function of
E + +D only. Eq. 20 thus represents a simple
mathematical condition for additivity of doses of
different types of radiation to yield an isoeffect.
This result is of extreme practical importance as it
establishes a quantitative criterion for the occurrence of synergistic isoeffects.
The condition of additivity, eq. 23, introduces
constraints among the coefficients of eq. 19. The
general form of # is obtained from eq. 20 as
and the constancy of Cpfor any values of D and E
necessarily demands that
Fig. 2. Partition coefficient surface for the synergism of X-rays
(the dose D is in Gy) and ultraviolet radiation (the dose E is
in J/m2) on Chinese hamster cells [ll]. The surface was drawn
according to eq. 24 with the parameter values [7,11]: (Y= 0.21
Gy-‘; /9 = 0.0266 Gy-‘; (I = 0 (J/m’); T = 0.0102 (J/m2)-2;
p - 0.016 Gy-’ (J/m2)-‘.
One sees that the surface is flat
only in the domain where the doses are relatively high. In this
domain the partition law, eq. 20, holds, the partition ccefficient cp is constant, and the doses that give an isoeffect are
additive.
ficients of eq. 24 has been reported for the case of
the synergism between X-rays and ultraviolet radiation on Chinese hamster cells [ll]. The corresponding partition coefficient surface is shown in
fig. 2. One sees that the surface is flat only at
relatively high doses, and therefore the partition
law, eq. 20, and hence the condition of additivity,
eq. 23, do not hold at low doses.
5. Discussion
These are necessary and sufficient conditions for
the validity of the partition law, eq. 20, and hence
of eq. 23. Substitution of these conditions into eqs.
17 and 18 shows that N + Q is a function of
E = $10 only.
A partition coefficient surface can be constructed from eq. 24 by plotting # vs. D and E.
Validity of the partition law, eq. 20, demands the
partition surface to be flat for arbitrary values of
D and E. Experimental determination of the coef-
The basic principles of linkage thermodynamics
[8,9] have been used to extend the LQ model [6,7]
in order to include linkage effects. We have shown
how the mutual interference of chemical binding
and radiation action, as well as the important
phenomenon of synergism, can be rationalized
with the help of simple considerations based on
the law of mass action. The phenomenological
treatment proposed here provides a more quanti-
E. Di Cera et at./ Linkage effects in a model for ceil survival after radiation
_I
97
.
.
.
n
I
-3’
-3
I
I
I
I
-1.5
I
I
0
I
I
I
1.5
3
Alfa
n
-3
-3
’ ’
’
’
-2
1’
’
’
’
1’
-1
’
’
’
1I
0
’
’
’
1’
1
’
’
’
1’
2
’
’
’
3
Alfa
Fig. 3. (a,b) Correlation plot for the u and B coefficients of eq. 1 in the text. The plot depicts the beat-fit values of n and j3 (in
standard deviation units) obtained by analysing 500 data sets simulated as described in the text, with a pseudorandom error of 10%
(a) and 50% (b). The values of the correlation coefficient are: -0.929 (a) and -0.939 (b),
98
E. Ri Cera et al/Linkage
eflecls in a model for cell survival after radiation
tative basis to extensive application
of the LQ
model to the analysis of experimental
datta. Of
particular relevance is the condition of additivity
of doses which follows directly from the partition
law, eq. 20. The importance of this law should be
brought out especially in connection with the therapeutical use of radiations, where it is critical to
establish alternative strategies of clinical treatment
that guarantee an isoeffect. More generally, the
theoretical approach to synergism and linkage effects proposed here can be most helpful in rationalizing experimental findings and making prcdictions on the effects of radiations on livin&.cells.
The approach considerably
broadens th& field of
applicability of the LQ model [7] and its biophysical relevance.
The strength of any approach to the effect of
radiation action on living cells cannot be exclusively based on the simplicity
and biophysical
significance
of its principles.
It necessarily also
stems from the extent to which one can resolve the
physical parameters
involved in the theoretical
treatment
from analysis of experimental
data.
Since this is a basic issue in the rigorous application of biophysical theories [20-221, we have addressed the problem of the resolvability of the a
and j3 coefficients of the basic eq. 1 of the LQ
model by extensive simulation studies. Data points
were simulated in the form usually taken experimentally, i.e., as the logarithm of S vs. the dose
D, using eq. 1 and adding a pseudorandom
error
to all points (40 per data set). The resolvability of
OLand fi was assessed by studying the distribution
of the best-fit values of these parameters collected _
from analysis of 500 simulated data sets. Due to
the linearity of the expression of the logarithm of
S with respect to D, the parameters a and /3 were
always normally distributed around the values used
in simulating the data. We” have also tested the
intrinsic correlation between 01 and p as a function of the experimental error. The distributions
of
the two parameters
are inversely correlated,
as
shown in fig. 3 for a typical case, but interestingly
enough the correlation
is barely affected by the
added pseudorandom
error in the range from 10
to 50%. This is the error range typically found
experimentally.
The thermodynamic
treatment
proposed here thus involves physical parameters
which can be readily extracted from analysis of
experimental
data. The intrinsic correlation
present between (Y and fi in the error range of a
typical experimental study does not seem to hinder
their resolvability,
and hence the validity of the
conclusions drawn.
References
5
6
7
8
9
10
11
12
13
14
15
16
17
1s
19
20
21
22
CA. Tobias, Radiat. Res. 104 (1985) S-77.
J. Kiefer, Radiat. Environ. Biophys. 27 (1988) 219.
H.S. Kaplan, Prcc. Natl. Acad. Sci. U.S.A. 55 (1966) 1442.
D.L. Dugle, C.J. Gillespie and J.D. Chapman, Proc. Natl.
Acad. Sci. U.S.A. 73 (1976) 809.
J.M. Clarkson and R.B. Painter, Mutat. Res. 23 (1974) 107.
K.H. Chadwick and H.P. Lee&outs,
Phys. Med. Biol. 18
(1973) 78.
K.H. Chadwick and H.P. Leenbouts, The molecular theory
of radiation biology (Springer, Berlin, 1981).
E. Di Cera, S.J. Gill and J. Wyman, Proc. Natl. Acad. Sci.
U.S.A. 85 (1988) 449.
E. Di Cera, S.J. Gill and J. Wyman, Proc. Nat]. Acad. Sci.
U.S.A. 85 (1988) 5077.
G.P. Raaphorst
and J. Kruuv, J. Radiat. Biol. 29 (1976)
493.
A. Han and M.M. Elkind, Int. J. Radiat. Biol. 31 (1977)
275.
D.L. Dugle and C.J. Gillespie, Molecular mechanisms of
repair of DNA (Plenum, New York, 1975).
ES. Ventsel, Teoria delle probabilita (Mir. Moskow, 1983).
W.K. Sinclair and R.A. Morton, Biophys. J. 5 (1965) 1.
W.K. Sinclair, Radiat. Res. 39 (1969) 135.
B.C. Cooke, E.M. Fielden and M. Johnson, Radiat. Res. 65
(1976) 152.
J. Wyman, Adv. Protein Chem. 4 (1948) 407.
E. Di Cera, M.L. Doyle, P.R. Connelly and S.J. Gill,
Biochemistry 26 (1987) 6494.
J.S. Haldane and J.L. Smith, J. Physiol. 22 (1897) 231.
M.L. Johnson and G.K. Ackers, Biophys. Chem. 7 (1977)
71.
E. Di Cera and S.J. Gill, Biophys. Chem. 29 (1988) 351.
S.J. Gill, P.R. Connelly,
E. Di Cera and C.H. Robert,
Biophys. Chem. 30 (1988) 133.