Download The origin of oncogenic mutations: where is the

Carcinogenesis vol.21 no.10 pp.1773–1776, 2000
COMMENTARY
The origin of oncogenic mutations: where is the primary damage?
Harald B.Steen
Department of Biophysics, Institute for Cancer Research, The Norwegian
Radium Hospital, Montebello, Oslo, N-0310 Norway
E-mail: [email protected]
Cancer is generally believed to arise from a single cell
which has become ‘initiated’ by mutation of a few crucial
genes, caused by random ‘hits’ in its DNA, a ‘hit’ being
an error in DNA replication or a reaction of the DNA with
free radicals or other chemical species of exogenous or
endogenous origin. It is not obvious how the epidemiological
data on cancer incidence can be interpreted within the
framework of this paradigm. For example, it cannot
account quantitatively for the age dependence of cancer
incidence, or for the fact that the incidence of cancer in
people with hereditary mutations in tumour suppressor
genes is much lower than expected, or for the observation
that while in some types of cancer, like colon and pancreas,
certain highly oncogenic mutations, such as that of TP53,
are prevalent, there is no significant increase in the
incidence of these cancers in people who carry the mutations by heredity. It is argued here that although mutations
in such genes appear to be of crucial importance in
carcinogenesis they may not be the rate limiting events in
common cancer. The epidemiological data are consistent
with the hypothesis that the rate limiting processes involve
large numbers of cells. Conceivably, the mutations directly
underlying neoplastic transformation may be the result of
a local collapse in the system of intercellular processes on
which the stability of the normal genotype and phenotype
depends, and thereby trigger a cascade of mutations, among
them the highly oncogenic ones. This local collapse may be
due to mutations of many different genes in many cells as
well as to other factors affecting the integrity of tissue.
Introduction
It is generally acknowledged that mutation has a key role in
carcinogenesis. During the last decade a number of genes have
been associated with carcinogenesis, either as oncogenes or
tumour suppressor genes. (In the following, these genes, i.e.
genes which may yield oncogenic mutations, will be denoted
‘oncogenic genes’.) Studies of hereditary cancer seem to show
that in some cases the loss of both alleles of a single tumour
suppressor gene may be sufficient to cause cancer, the classical
case being retinoblastoma (Rb) (1). In general, however, the
connection between mutation and cancer appears to be more
complex.
Epidemiological studies are consistent with the idea that
cancer is due to a small number of discrete events. Armitage
and Doll (2) calculated from the age dependence of the
mortality of several of the most common types of cancer that
© Oxford University Press
the number of such events is five to seven. A similar, more
recent study, based on more extensive statistics and covering
a broader spectrum of cancers, found the majority of cancers
in the range four to eight events with extremes of three and
12 (3). It should be emphasized that in these studies the
conclusions are based on the assumption that the oncogenic
events are rate limiting with regard to cancer incidence. Genetic
and molecular studies have identified these events as mutations
within a limited set of genes, i.e. the oncogenic genes. Thus,
there is a general consensus that cancer arises from a single,
neoplastic cell, and that the reason for the initiation of this
cell is mutation of a few crucial genes. The oncogenic mutations
may be the result of free radicals or other chemical species,
of exogenic or endogenic origin, reacting with the DNA of
the cell, i.e. by ‘direct hits’ causing irreparable malfunction of
the respective genes. They may also result from replication
errors. Such initiating events are assumed to hit essentially at
random, although some cells may be more exposed than others,
and genes may differ somewhat in their susceptibility to
mutation. In any case it is taken for granted that the primary
damage causing mutation occurs in the cell which eventually
becomes the origin of the malignant clone. In other words,
that the rate limiting processes in the initiation of cancer take
place in that very cell. It is essential in discussing these matters
to realise that although mutations of oncogenic genes may be
crucial in carcinogenesis, they are not necessarily rate limiting,
as is a tacit assumption in many studies and considerations of
cancer causation.
Although a great variety of mutagens, including chemical
substances and different types of radiation, have been identified,
the mechanisms leading to the oncogenic mutations are not
fully understood. The currently accepted model implies that
the oncogenic mutations, which lead to the initiation of
carcinogenesis, are essentially independent of interaction with
other cells, although the subsequent development of the
malignant phenotype, i.e. the ‘promotion phase‘, may depend
on a variety of extracellular factors, such as hormones, immunological compatibility, stroma interactions, etc. Thus, the hypothesis that cancer originates from a single cell which has
become ‘initiated’ by mutation of a few crucial genes caused
by random ‘hits’ in its DNA, has become a paradigm of
current cancer research. In the following it is shown that this
paradigm does not seem to be in accordance with epidemiological data. These data rather suggest that processes involving
intercellular interactions may be rate limiting also in the
initiation phase of carcinogenesis.
In the present discussion I am concerned with the bulk of
cancer cases, that is what may be called common cancer,
which appear to arise spontaneously with no obvious reason,
although exogenic and/or endogenic carcinogens are likely to
be involved. My arguments do not necessarily apply to cases
associated with heavy exposure to carcinogenic radiation or
chemicals.
1773
H.B.Steen
The model
As a framework for this discussion we consider the following
model which contains the above paradigm as a special case.
Thus, we assume a system, i.e. a tissue or organ, consisting
of Nc compartments, each containing n cells, each cell with c
targets, e.g. oncogenic genes. We assume further that this
system is being hit at random by discrete, independent events
at a rate s per target and unit time, and that a certain number,
x, of such initiating events among the n·c targets in any one
of the Nc compartments will result in a neoplastic cell which
eventually may develop into cancer. In terms of the current
paradigm each compartment is one cell, i.e. n ⫽ 1, s is the
rate of mutation and c is the number of oncogenic genes per
cell. (A gene in this context may be just one allele of a gene.)
Assuming that the initiating events occur randomly we apply
Poisson statistics to calculate the probability, p(x), that one
compartment shall have x or more hits in the c targets within
a period of time, t:
⬁
p(x) ⫽
Σ (µ e–µ)/x!,
x
µ ⫽ ncst
1
x
where s is the probability of one hit per target and unit time.
Since µ ⬍⬍ 1, eqn 1 reduces to:
p(x) µ µxe–µ/x! µ µx (1 – µ)/x! µ µx/x!
2
The probability that none of the Nc compartments will
experience x or more hits is:
R(x) ⫽ [1 – p(x)]Nc µ 1 – Nc p(x)
3
The approximation being valid since Nc·p(x) ⬍ 1. Hence, the
probability, P(x), that any one of the Nc compartments will
experience x or more hits, i.e. that an initiated cell will occur
in the system is:
P(x) ⫽ 1 – R(x) µ Nc µx/x! ⫽ Nc(ncst)x/x!
4
To see the consequences of the current paradigm we put
n ⫽ 1, which is to say that each compartment contains just
one cell, and assume initially that c ⫽ x, which is to say that
oncogenic transformation requires a unique combination of
mutations, as seems to be the case for retinoblastoma. (This
assumption is probably not valid for most types of cancer.
However, it simplifies the calculations and, as we shall see, it
makes essentially no difference to the conclusions as long as
c is not a very large number.) If we apply this to a hypothetical
organ containing Nc ⫽ 1⫻1010 cells with an oncogenic potential
(i.e. roughly the number of cells in 10 g tissue), P(x) ⫽ 0.03,
s ⫽ 3⫻10–7/gene/cell/year, and t ⫽ 50 years, we calculate
from eqn 4 that x µ 2.4. (If we assume c ⫽ 5x the estimate
of x increases to 2.7.) This estimate is rather on the high side
since it is assumed that every initiated cell, that is every cell
with x oncogenic mutations, gives rise to an observable cancer.
If the promotion phase is rate limiting at all the value of x
must be smaller. But not necessarily very much; for example,
if one in every 10 initiated cells gives rise to cancer, x is
reduced only from 2.4 to 2.2, and if it is one in 100, x µ 2.0.
Of course, the values of the various parameters that go into
this calculation are order of magnitude estimates which may
vary significantly between different organs, etc. However, the
value of x is not very sensitive to variations in these parameters.
For example, changing the value of P(x) or Nc by a factor of
10 alters x by ⬍10%, while a similar change of s, c or t gives
1774
a change in x of ~20%. Hence, we see that, even if one takes
into account that the estimates of the basic parameters, i.e.
rate of mutation, the number of potentially oncogenic cells,
etc., may be off by an order of magnitude or more, and that
the promotion phase may be rate limiting, the current paradigm
leads to a number of primary oncogenic events which is
incompatible with the epidemiological data referred to above
(2,3).
On the other hand, if we accept the average epidemiological
estimate of x ⫽ 6 the current paradigm leads to an excessively
high rate of mutation. Thus, with n ⫽ 1 eqn 4 yields a rate of
mutation s µ 1.2⫻10–4/gene/cell/year. This is comparable with
values achievable in in vitro assays with highly potent mutagens
applied in concentrations significantly reducing cell survival,
and therefore seems unrealistically high as a general rate of
mutation in humans. Estimates of in vivo rates of ‘spontaneous’
mutation fall in the region 1⫻10–7 to 5⫻10–6/gene/cell/year
(4,5). Alternatively, the value of c would have to be quite
large, i.e. in the order of c µ 2400, which is to say that
virtually every active gene of the cell would be oncogenic,
again in contradiction to the current paradigm. Hence, this
paradigm (eqn 4 with n ⫽ 1) cannot account for the age
dependence of the incidence of most types of common cancer.
The discrepancy may be resolved if we assume that n ⬎⬎
1, which is to say that each compartment contains not just
one, but many cells. For example, if we take x ⫽ 6, in
accordance with the epidemiological data, and the other
parameters as above, eqn 4 yields n·c µ 2400. In calculating
that number we assume that the number of compartments equals
that of cells (Nc ⫽ N), which implies that the compartments are
overlapping, i.e. that every cell interacts with many others.
(Assuming non-overlapping compartments, i.e. Nc ⫽ N/n,
yields n·c µ 8000.) Hence, it appears that the data are
consistent with a model which assumes that the events leading
to a neoplastic cell need not to be confined to that very
cell, but may occur within a relatively large number of its
neighbours, and/or that the number of targets, i.e. oncogenic
genes, is much larger than indicated by molecular studies.
Another serious discrepancy between the current paradigm
and the epidemiological data appears when we consider the
increase of cancer incidence in people carrying a hereditary
mutation in one of the c oncogenic genes, such as Li–Fraumeni
patients. From eqn 4 we may calculate the magnitude of this
increase, that is the risk ratio Q, assuming that every cell has
a handicap of one event, i.e. mutation in one of its oncogenic
genes, so that the number of events required to initiate cancer
is reduced by one:
Q ⫽ P(x – 1)/P(x) µ x/µ ⫽ x/ncst
5
The current paradigm, that is, taking: n ⫽ 1, and the other
parameters as above, gives: Q µ 7⫻104. Allowing for the age
difference, this value is in accordance with that observed for
retinoblastoma in children (1). However, it is very much higher
than what is usually observed with hereditary mutations linked
with common types of cancer, such as mutations in the TP53
gene or the BRCA1 gene, for which it is typically within a
factor of 100 (6), and for most cancers much below that.
Again, the discrepancy can be resolved by assuming that
n·c is large, i.e. in the order of several thousand. For example,
assuming Q ⫽ 100 and the other parameters as above, eqn 5
yields n·c ⫽ 4000. Hence, again we are led to conclude that
the primary events may take place in other cells than the
one which eventually becomes initiated, which implies that
The origin of oncogenic mutations
intercellular interactions may have a crucial role in the etiology
of common cancer.
It may be argued that an alternative explanation for the
relatively low incidence of cancer in people carrying hereditary
oncogenic mutations may be that these mutations cause only
a moderate increase in the general rate of mutation, i.e. a
degree of genomic instability, for example mutations of genes
controlling the fidelity of DNA replication, such as believed
to be one function of TP53, and the mismatch repair genes,
or genes for enzymes which metabolize certain carcinogenic
molecules, etc. It turns out, however, that if a mutation of one
of the oncogenic genes has such an effect on genomic stability
the result would be rather the opposite. To illustrate this, we
rewrite eqn 4 to allow for the assumption that different genes
may have different rates of mutation:
P(x) ⫽ Nc·(nct)x·(s1·s2· . . . ·sx)/x!
6
and the similar expression for the case with a hereditary
mutation in one of these genes (otherwise with mutation rate
s1) so that the mutation rate of the other genes are enhanced:
P’(x – 1) ⫽ Nc·(nct)x–1·(s’2· . . . ·s’x)/(x – 1)!
7
and take the ratio between the two:
Q’ ⫽ P’(x – 1)/P(x) ⫽ (x/ncts1)·(s’2·s’3· . . . ·s’x)/
(s2·s3· . . . ·sx)
8
Since s’ ⬎ s, due to the genomic instability caused by the
hereditary mutation, the last ratio is larger than unity and the
value of Q’, therefore even larger than that calculated from
eqn 5. The same argument holds true if the mutation which
increases the probability of subsequent mutations occurs later
in the sequence, although its effect will be smaller. Hence, if
TP53, the mismatch repair genes (7), and other tumour
suppressor genes associated with genomic stability, are to be
included in the set of the c oncogenic genes, and that
assumption is indeed the basis for much of the current
interest in this area, the possible effect of mutation of these
genes on genomic stability cannot account for the observations; in contrast, it makes the discrepancy with the current
paradigm even larger. This argument also applies to the case
where the first, or one of the first, mutations gives rise to a
clone with increased proliferation, so that partly initiated
cells multiply faster than normal. Thus, in the above model
(equations 6–8) such an increase is equivalent to an increase
in the rate of mutation, s, and therefore would yield the same
result as increased genomic instability, namely to increase the
discrepancy with the current paradigm rather than reduce it.
Discussion
The present model encompasses, as a special case, the current
paradigm, namely that cancer originates from a single cell
which has become ‘initiated’ by mutation of a few crucial
genes caused by random ‘hits’ in its DNA. The above calculations show that this paradigm does not seem to be in accordance
with the epidemiological data. On the other hand, compatibility
with these data is obtained if it is assumed that the primary
events may take place not only in the cell which becomes
initiated, but within a large number of cells, and affecting a
larger number of genes.
The low number of primary events, i.e. the value of x,
which results when we apply the rate of spontaneous mutation
in eqn 4, is not only in disagreement with the estimates based
on epidemiological data, it is also in contrast with the multiple
mutations and chromosome aberrations seen in most cancers,
although it is not known at which stage in the process of
malignant transformation this damage occurs. An explanation
highlighted by Loeb (8) is that the neoplastic cell acquires a
‘mutator phenotype’ early in its development. There is ample
evidence of such genomic instability in cancer, and several
genes which confer such instability upon mutation have been
identified (9). Mutation of these genes can be a heritable trait,
which is associated with a higher than normal incidence of
some types of cancer. As shown above, however, to bring this
hypothesis in agreement with the epidemiological data we
have to assume that the primary events are distributed over
many cells and possibly many genes.
One interpretation of the observation that the relative risk
ratio for people with a hereditary mutation in a tumour
suppressor gene is much lower than that calculated from eqn
5, is that that mutation is not a primary rate limiting factor in
the incidence of common cancers. However, that is not to say
that this mutation cannot be crucial in the etiology of such
cancers. Thus, if the rate limiting event is a process which
raises the rate of oncogenic mutation radically, it will be that
process, and not the ensuing oncogenic mutations, which is
rate limiting, although the oncogenic mutations may still be a
prerequisite for the development of the neoplastic cell. This
may explain why hereditary mutations in tumour suppressor
genes, like TP53, do not cause significant increase of the
incidence of many cancers, like colorectal and pancreatic,
although mutations of these genes are very common in such
cancers, and appear to play a definite role in their etiology
(10,11).
What, then, could be the nature of the rate limiting events
which induce the oncogenic mutations that give rise to the
neoplastic cell? The above considerations suggest that, typically, such events involve large numbers of cells, which is
to say that collective, intercellular processes, or rather the
interruption of such processes, may be essential in the initiation
of carcinogenesis of common cancer in man. Numerous in vitro
studies demonstrate that a variety of extracellular signals
contribute to the control of gene expression, differentiation
and proliferation. This variety includes diffusive chemical
factors, like hormones and growth factors, electrochemical
interactions, notably through Ca2⫹-mediated via gap junctions,
and mechanical and structural interactions mediated through
the anchorage of cells to the extracellular matrix and to
adjacent cells (12). In this context it may be relevant that at
least one tumour suppressor gene product, i.e. that of the APC
gene, is extra-nuclear and appears to have a crucial role in
transmittance of signals from the cell exterior and into the
nucleus (13). Moreover, these mechanisms may be coupled.
For example, it has been shown that mechanical tension applied
to a cell in a confluent cell culture may trigger a wave of Ca2⫹
which may be transmitted over distances of many cell diameters
(14,15). Vice versa, Ca2⫹ is known to affect the polymerization/
depolymerization of cytoskeleton components responsible for
the mechanical forces exerted by cells on their surroundings.
Such coupling provides a basis for an extended network of
feedback loops. In more general terms, the organism represents
a highly complex network of information flow where proper
function and stability of each element is dependent on a
multitude of appropriate signals from many other elements.
Thus, it may be argued that essentially every gene is part of
this network, on the grounds that any gene product which is
1775
H.B.Steen
not essential to the organism, i.e. takes part in the network,
would disappear in evolution. Although this may be an oversimplification, it constitutes a rationale for the assumption that
a large number of genes are involved in the maintenance of
the integrity of a tissue which in turn is a condition for the
maintenance of genomic stability. A consequence of this
assumption is that a mutation in any gene of any cell may,
however little, contribute to a degradation of the integrity of
the system until eventually it reaches a point where it is no
longer able to keep every cell in check. The result of such
degradation could be a local collapse of the system, leading
to a drastic increase of the probability of an initiating oncogenic
event in that locality. Thus, the rate limiting events may be
mutations in virtually any of the genes which contribute to
the integrity of the tissue, rather than direct ‘hits’ in the
individual oncogenic genes. This is not to say that cancer does
not originate from a single cell, but that the malignant
transformation of that cell may be a secondary result of many
mutations and possibly other types of damage in a large
number of neighbouring cells. Thus, not only mutation, but
any type of damage that may degrade the cellular environment
over a significant period of time, may increase the probability
of an initiating event. This may be a rationale for the carcinogenic effect of some of the carcinogens which do not seem to
be mutagenic.
Cancer cells are generally characterized not only by
mutations of specific oncogenic genes, but a diverse spectrum
of chromosome aberrations (16). Malignant tumours display a
variety of major deletions and amplifications, each of which
affects stretches of chromatin typically containing hundreds
or thousands of genes. Although there is a tendency to
clustering on certain chromosomes, these aberrations, on
average six to eight per cell in cancers like breast (17), nonHodgkin’s lymphomas (18) and colon (19), are generally
spread over the entire genome, with significant variation
between individual tumours and also tumours of the
same type. Such data demonstrate both the commonness of
chromosome aberrations in cancers and the essentially random
nature of the processes by which they occur. Conceivably, the
‘local collapse’ may lead to a significant increase in the
probability of chromosome aberrations. Aberrations are
believed to destabilize the chromatin and further increase
the probability of new aberrations (20,21). Thus, the ‘local
collapse’ may trigger a cascade of mutations, among them the
highly oncogenic ones.
If the present conclusion, namely that collective effects play
a significant role in the initiation phase of carcinogenesis, can
be substantiated it implies new perspectives both in prevention
and in treatment of cancer.
References
1. Knudson,A.G. (1971) Mutation and cancer: A statistical study of
retinoblastoma. Proc. Natl Acad. Sci. USA, 68, 820–823.
1776
2. Armitage,P. and Doll,R. (1954) The age distribution of cancer and a
multistage theory of carcinogenesis. Br. J. Cancer, 8, 1–12.
3. Renan,M.J. (1993) How many mutations are required for tumorigenesis?
Implications from human cancer data. Mol. Carcinog., 17, 139–146.
4. Bridges,B.A., Cole,J., Favor,J., Glickman,B.W., Mohrenweiser,H.,
Sankaranarayanan,K. and Skopek,T.R. (1994) Spontaneous mutation and
its place in risk assessment for chemical mutagens. Report of an ICPEMC
committee. Mutat. Res., 304, 3–11.
5. Simpson,A.J.G. (1997) The natural somatic mutation frequency and human
carcinogenesis. Vande-Woude, G.F. and Klein, G. (eds) Advances in
Cancer Research. Academic Press, New York, 209–240.
6. Bishop,D.T. (1996) Genetic predisposition to cancer: an introduction.
Eeles,R.A., Ponder,B.A.J., Easton,D.F., Horwich,A. and McVie,G. (eds)
Genetic Predisposition to Cancer. Chapman & Hall Medical, London,
pp. 3–15.
7. Loeb,L.A. (1994) Microsatellite instability: marker of a mutator phenotype
in cancer. Cancer Res., 54, 5059–5063.
8. Loeb,L.A. (1991) Mutator phenotype may be required for multistage
carcinogenesis. Cancer Res., 51, 3075–3079.
9. Lengauer,C., Kinzler,K.W. and Vogelstein,B. (1998) Genetic instabilities
in human cancers. Nature, 396, 643–649.
10. Lundin,J., Nordling,S., von Boguslawsky,K., Roberts,P.J. and Haglund,C.
(1996) Prognostic value of immunohistochemical expression of p53 in
patients with pancreatic cancer. Oncology, 53, 104–111.
11. Breivik,J., Lothe,R.A., Meling,G.I., Rognum,T.O., Børresen-Dale,A.L. and
Gaudernack,G. (1997) Different genetic pathways to proximal and distal
colorectal cancer influenced by sex-related factors. Int. J. Cancer, 74,
664–669.
12. Weaver,V.M., Petersen,O.W., Wang,F., Larabell,C.A., Briand,P., Damsky,C.
and Bissell,M.J. (1997) Reversion of malignant phenotype of human breast
cells in three-dimensional culture in vivo by integrin blocking antibodies.
J. Cell Biol., 137, 231–245.
13. Ilyas,M. and Tomlinson,I.P. (1997) The interactions of APC, E-cadherin
and beta-catenin in tumour development and progression. J. Pathol., 182,
128–137.
14. Sanderson,M.J., Charles,A.C., Boitano,S. and Dirksen,E.R. (1994)
Mechanisms and function of intercellular calcium signaling. Mol. Cell.
Endocrinol., 98, 173–187.
15. Sneyd,J., Keizer,J. and Sanderson,M.J. (1995) Mechanisms of calcium
oscillations and waves: a quantitative analysis. FASEB J., 9, 1463–1472.
16. Mitelman,F., Mertens,F. and Johansson,B. (1997) A breakpoint map of
recurrent chromosomal rearrangements in human neoplasia. Nature Genet.,
15, 417–474.
17. Tirkkonen,M., Tanner,M., Karhu,R., Kallioniemi,A., Isola,J. and
Kallioniemi,O.P. (1998) Molecular cytogenetics of primary breast cancer
by CGH. Genes Chromosomes Cancer, 21, 177–184.
18. Monni,O., Oinonen,R., Elonen,E., Franssila,K., Teerenhovi,L., Joensuu,H.
and Knuutila,S. (1998) Gain of 3q and deletion of 11q22 are frequent
aberrations in mantle cell lymphoma. Genes Chromosomes Cancer, 21,
298–307.
19. Angelis,P.M., Clausen,O.P.F. and Stokke,T. (1999) Chromosomal gains
and losses in primary colorectal carcinomas detected by CGH and their
associations with tumor DNA ploidy, genotypes and phenotypes. Br. J.
Cancer, 80, 526–535.
20. Cram,L.S., Bartholdi,M.F., Ray,F.A., Travis,G.L. and Kraemer,P.M. (1983)
Spontaneous neoplastic evolution of chinese hamster cells in culture:
multistep progression of karyotype. Cancer Res., 43, 4828–4837.
21. Kuukasjarvi,T., Karhu,R., Tanner,M., Kahkonen,M., Schaffer,A.,
Nupponen,N., Pennanen,S., Kalliomiemi,A., Kalliomiemi,O.-P. and Isola,J.
(1997) Genetic heterogeneity and clonal evolution underlying development
of asynchronous metastasis in human breast cancer. Cancer Res., 57,
1597–1604.
Received December 16, 1999; revised May 17, 2000; accepted June 21, 2000