Download Tumors with microsatellite instability: many mutations, targets and

Oncogene (2003) 22, 2223–2225
& 2003 Nature Publishing Group All rights reserved 0950-9232/03 $25.00
www.nature.com/onc
COMMENTARY
Tumors with microsatellite instability: many mutations, targets and
paradoxes
Manuel Perucho
The Burnham Institute, La Jolla Cancer Research Center, 11099 North Torrey Pines Road, La Jolla, CA 92037, USA
Oncogene (2003) 22, 2223–2225. doi:10.1038/sj.onc.1206580
Tumors of the microsatellite mutator phenotype (MMP)
accumulate hundreds of thousands of somatic clonal
mutations, preferentially in simple repeats or microsatellites (Ionov et al., 1993). Among this multitude of
irrelevant mutations in ‘junk’ DNA, quite a few also hit
genomic coding regions with potential impact on
phenotype. The issue is then how to sort out cause
from consequence and mutations that contribute from
mutations that are inconsequential to tumorigenesis.
In this issue of Oncogene, Woerner and colleagues
make a good effort in this direction, based on a metaanalysis of the mutations in genes with simple repeats
reported so far in the literature. The list is extensive and
includes more than 100 genes in colon cancer, and near
50 in gastric and endometrial cancers, which are the
three human cancers with the highest incidence of
microsatellite instability (MSI). The work is based
on more than 60 publications during the past few
years, which is a good barometer for the interest in this
area.
The study is intended to distinguish, among this glut
of mutated genes, those whose mutations during tumor
progression are under positive selection (i.e. relevant,
functional mutations) and those that are not (i.e
irrelevant, neutral mutations). The statistical approach
is based on stratification of mutation frequency. Some
genes showing high mutation frequency are thus
categorized as ‘common targets’ for the MMP in colon,
gastric and endometrial cancer. The list includes some
familiar names, such as TGFbRII and BAX, and other
newcomers. The effort is laudable, in the same line as
previous reports by Duval, Iacopetta, Hamelin and
colleagues (Duval et al 2002a; Duval and Hamelin,
2002) that attempt to find objective criteria for
determining mutation functionality.
The topic needs to be put into the right context. The
first thought to come to the mind of a hypothetical
unprejudiced reader could be to question the necessity to
discern the functionality or neutrality of many novel
mutated cancer genes in the genesis of tumors with MSI.
For instance, colon cancer is well known to involve the
mutational activation and inactivation of the prototypical cancer genes K-ras oncogene, and APC and p53
tumor suppressors (Kinzler and Vogelstein, 1995). Then,
why colon cancers with MSI would need additional
mutated cancer genes? Would it not be more logical to
conclude that mutations in microsatellite sequences,
regardless for their coding or noncoding location, are all
inconsequential for tumorigenesis in colon tumors with
DNA mismatch repair (MMR) deficiency? If they have
already become tumors in the first place, they might
have done so because of the obligatory APC, K-ras and
p53 oncogenic mutations (Kinzler and Vogelstein, 1995).
Would it not be therefore a non-productive and
distractive task to scrutinize the 100 plus novel target
cancer genes in these tumors?
The answer to such questions is ‘certainly not’. This is
because of the paradoxically low incidence of mutations
in APC, K-ras and p53 in colon tumors of the MMP
(Ionov et al., 1993; Kim et al., 1994; Konishi et al., 1996;
Perucho, 1999), the very same tumors that may cram
millions of somatic mutations in other DNA sequences.
Indeed, a first level of explanation for this paradox is
precisely the existence of these MSI-specific cancer
genes. Genes never observed mutated in tumors before
are now netted in the handful in MSI-positive tumors,
because these tumors exhibit a more than two orders of
magnitude increase in slippage errors in simple repeats.
The first such target gene described was TGFbRII
(Markowitz et al., 1995), which has been found to be
mutated in a vast majority of gastrointestinal tumors
with MSI. It is no surprise that it is classified in the
‘functional’ common target category by Woerner et al.
(in press) statistical approach.
However, it is difficult to distinguish between relevant
and irrelevant mutations in MSI-positive tumors, and
the statistical approach of Woerner et al. is not seamless,
as they acknowledge. While meta-analyses are intended
to erase spurious fluctuations of single observations,
some genes included in the study have been analysed
only by one group. This leads to true incidence
uncertainty. As the procedure relies on the localization
of the data points inside or outside an inconclusive
bracket zone, slight fluctuation in mutation frequency
transmutes the status of some target genes from
functional to non-committal and vice versa. For
instance, one of the newly identified common target
genes is HT001, for which there is no evidence of
involvement in neoplasia. The gene exhibits the highest
mutation incidence (17 mutations in 20 tumors, 85%),
Commentary
2224
but just adding one or two more negative tumors would
‘demote’ the gene to the ambiguous class by decreasing
the frequency below the diagnostic boundary.
On the other hand, genes for which there is evidence
for the functionality of mutations, such as BAX (Gil
et al., 1999; Ionov et al., 2000), MSH6 (Baranowskaya
et al., 2001), Axin (Liu et al., 2000) or IGFRII (Souza
et al., 1999), do not make the cutoff line for relevancy in
some or all of the MSI-positive tumors. Therefore, these
studies fall under the old maxim that if observations rely
on statistical validation, it would suggest an urgent need
to perform a better experiment. Of course, there is the
argument that the approach formally forbids the
conclusion that genes within the inconclusive area are
not functional (the method only establishes that they do
not escape the category of doubtful targets). However,
this would amount to little more than statistical
rhetorical sophistry. Nevertheless, while not perfect,
the approach is useful.
Thus, Woerner et al. describe not only how to spot
‘true’ target genes, but also genes whose mutations may
be under negative selection during tumor progression.
This is important because it addresses the problematic
issue of establishing the relevancy of the absence of an
event rather than its presence. It could be viewed as a
situation equivalent to the insightful Sherlockian conclusion (in the ‘Adventure of Silver Blaze’) that the
relevant incident was that the dog did not bark in the
night (when the racing horse disappeared). For absence
of evidence is sometimes evidence of absence. Thus, a
couple of genes (CHD2 and RFC3) involved in genome
integrity are categorized as being under negative
selective pressure in colorectal tumorigenesis because
of their low mutation incidence. This is interesting,
although some other genes, such as DNA polymerase a,
that exhibit a conspicuous absence of mutations
(Yamamoto et al., 1997) are not listed in this category.
This is because Pola only has a repeat of eight
nucleotides, and the approach by Woerner et al.
excludes genes that have repeats shorter than 10
nucleotides from being unambiguously classified as
subject to negative selective pressure.
Moreover, two non-coding repeats are listed as being
under negative selection, which is somewhat odd. The
same kind of oddity would be to include non-coding
repeat mutations under positive selection. This could
have happened if some of the non-coding repeats
reported to display a very high mutation frequency
(Zhang et al., 2001) had not been excluded from the
meta-analysis, as they were shown to be length polymorphisms rather than somatic mutations (Duval et al.,
2002b; Suzuki et al., 2002). However, it is quite possible
that MSI-specific mutations in non-coding sequences
may also contribute to the manifestation of the
malignant phenotype, for instance if they occur in gene
regulatory regions that control levels of expression. It is
tempting to predict that the list of such genes with
putative roles in tumors with MSI will grow in the
future, perhaps to the size of the list of genes with
mutations in coding repeats. However, it is not clear
whether the deletion of a single nucleotide in a 3’ UTR
Oncogene
homopolymeric tract of 10 adenines could have a clear
impact in gene expression (for instance, by altering the
half-life of the messenger RNA) to deserve to be the
target of a strong negative selection during tumor
progression.
Thus, it appears more plausible that mutations (or
lack of mutations) in some repeats are actually found (or
not found), not because they are under positive (or
negative) selection but because of abnormally high (or
low) repeat mutation frequencies. Microsatellites of an
identical number of repeat units may exhibit different
mutability, regardless of coding capabilities, because of
the influence of the surrounding sequences in replication
fidelity (Zhang et al., 2001; Duval et al., 2002a, b; Suzuki
et al., 2002).
Figure 1 of this editorial complies the data from
Woerner et al. (in press), including genes with repeats of
7, 8, 9 and 10 units, ordered by decreasing mutation
frequency. Not all these genes found mutated in cancer
of the MMP ought to be relevant for the manifestation
of the malignant phenotype by these tumor cells.
Especially considering that the genes represented are
far from comprising the entire set of genes with repeats
in the human genome. In other words, many of these
mutations are probably inconsequential. However, the
figure also highlights the intrinsic difficulty in deciding
the relevancy of mutations based on mutational
frequency alone. Where to place the relevancy/irrelevancy cut-off line? No matter which statistical method is
used, any such decision would always be tainted by a
great deal of arbitrariness.
The mutation incidence of microsatellite sequences in
an MMR-deficient background thus can be very high or
very low (Figure 1). While genes with mutations found
with high incidence have already a good lead to merit
inclusion into the category of ‘relevant’, genes with
infrequent mutations may not be regarded as irrelevant.
This is because these tumors accumulate so many
mutations that disruption of cell growth and survival
regulation can be accomplished in different tumors by
mutations in different genes of the same signaling
networks (Suzuki et al., 2002). This may lead to a lower
incidence of any particular gene, relative to tumors
without mutator phenotype, where very rare mutations
are funneled into a few cancer genes with strong
oncogenic potential. The first paradox posed by the
low incidence of mutations in APC, K-ras or p53 in
tumors of the MMP can be thus rationalized. In
Figure 1 Genes with coding repeats.
Commentary
2225
conclusion, in this scenario, the relevance of microsatellite-specific mutations in MSI-positive tumors can
be proven only when there is supporting evidence for
functionality, regardless of mutation incidence.
However, all these complications are in fact minor
compared to the more fundamental issue of the allelic
status of these mutated target genes. There is no doubt
that biallelic mutations in genes with proven roles in cell
growth or survival have an impact in cell phenotype,
regardless of incidence. However, another paradox of
tumors with MSI is that while biallelic mutations in
irrelevant (i.e. non-coding) sequences are bountiful
(>105, Ionov et al., 1993), mutations in relevant (i.e.
coding) sequences (i.e. those of Figure 1) are often
monoallelic (see for instance Markowitz et al., 1995;
Rampino et al., 1997; Yamamoto et al., 1997, 2000). But
to delve into this issue would take much more space
than allowed in this commentary.
References
Duval A and Hamelin R. (2002). Cancer Res., 62, 2447–2454.
Duval A, Reperant M, Compoint A, Seruca R, Ranzani GN,
Iacopetta B and Hamelin R. (2002a). Cancer Res., 62, 1609–
1612.
Duval A, Reperant M and Hamelin R. (2002b). Oncogene, 21,
8062–8066.
Gil J, Yamamoto H, Zapata JM, Reed JC and Perucho M.
(1999). Cancer Res., 59, 2034–2037.
Ionov Y, Peinado MA, Malkhosyan S, Shibata D and Perucho
M. (1993). Nature, 363, 558–561.
Ionov Y, Yamamoto H, Krajewski S, Reed JC and Perucho
M. (2000). Proc. Natl. Acad. Sci. USA, 97, 10872–10877.
Kim H, Jen J, Vogelstein B and Hamilton SR. (1994). Am J
Pathol, 145, 148–156.
Konishi M, Kikuchi-Yanoshita R, Tanaka K, Muraoka M,
Onda A, Okumura Y, Kishi N, Iwama T, Mori T, Koike M,
Ushio K, Chiba M, Nomizu S, Konishi F, Utsunomiya J
and Miyaki M. (1996). Gastroenterology, 111, 307–317.
Liu W, Dong X, Mai M, Seelan RS, Taniguchi K,
Krishnadath KK, Halling KC, Cunningham JM, Boardman
LA, Qian C, Christensen E, Schmidt SS, Roche PC, Smith
DI and Thibodeau SN (2000). Nat. Genet., 26, 146–147.
Markowitz S, Wang J, Myeroff L, Parsons R, Sun L,
Lutterbaugh J, Fan RS, Zborowska E, Kinzler KW and
Vogelstein B. (1995). Science, 268, 1336–1338.
Perucho M. (1999). Cancer Res., 59, 249–256.
Rampino N, Yamamoto H, Ionov Y, Li Y, Sawai H, Reed JC
and Perucho M. (1997). Science, 275, 967–969.
Souza RF, Wang S, Thankar M, Smolinski KN, Yin J, Zou
TT, Kong D, Abraham JM, Toretsky JA and Meltzer SJ.
(1999). Oncogene, 18, 4063–4068.
Suzuki K, Dai T, Suzuki I, Dai Y, Yamashita K and Perucho
M. (2002). Cancer Res., 62, 1961–1965.
Woerner SM, Benner A, Sutter C, Schiller M, Yuan P, Keller
G, Bork P, von Knebel Doeberitz M and Gebert JF.
Oncogene, (in press).
Yamamoto H, Gil J, Schwartz S, Perucho M. (2000). Cell
Death Differ., 7, 238–239.
Yamamoto H, Sawai H and Perucho M. (1997). Cancer Res.,
57, 4420–4426.
Zhang L, Yu J, Willson JK, Markowitz SD, Kinzler KW and
Vogelstein B. (2001). Cancer Res., 61, 3801–3805.
Oncogene